JP6998823B2 - Multi-channel objective evaluation device and program - Google Patents

Description

The present invention relates to a multi-channel objective evaluation device and a program for objectively evaluating the quality of a multi-channel acoustic signal used in a multi-channel acoustic system having more than two channels.

Conventionally, in a multi-channel acoustic system, a method of evaluating the quality of an acoustic signal has been known. For example, as a method for subjectively evaluating the quality of an acoustic signal, a subjective evaluation method for an acoustic system with little deterioration including a multi-channel acoustic system is described in ITU-R Recommendation BS. It is defined in 1116-3 (see, for example, Non-Patent Document 1).

On the other hand, ITU-R Recommendation BS. The objective evaluation method for objectively measuring the sound quality corresponding to the subjective evaluation performed in accordance with 1116-3 is ITU-R Recommendation BS. 1387-1 (see, for example, Non-Patent Document 2). This ITU-R recommendation BS. The objective evaluation method defined in 1387-1 is called a PEAQ (Perceptual Evaluation of Audio Quality) objective sound quality measurement method.

The PEAQ objective sound quality measurement method is realized by a standardized algorithm for objectively measuring the quality of acoustic signals, and uses an auditory model that reflects the perceptual characteristics of the human ear and a recognition model that has a neural network structure. It seeks an objective evaluation value. Details will be described later.

In general, it is not realistic to perform a subjective evaluation on all sound sources because a large number of subjects, a large amount of time and effort are required to perform a highly reliable subjective evaluation. Therefore, the parameters used for the subjective evaluation are selected by performing the objective evaluation in advance.

However, the aforementioned ITU-R Recommendation BS. The objective evaluation method defined in 1387-1 is a method applied to a one-channel or two-channel acoustic system. Therefore, this objective evaluation method cannot be used for a multi-channel acoustic system (see, for example, Non-Patent Document 3) having more than two channels such as 22.2 ch (channel).

Therefore, a method for objectively evaluating the quality of a multi-channel acoustic signal in a multi-channel acoustic system having more than two channels has been proposed (see, for example, Non-Patent Document 4). In this method, the head related impulse response (HRIR) is convoluted with the original sound and the deteriorated sound of the multi-channel acoustic signal, respectively, and converted into a two-channel signal for objective evaluation.

Rec. ITU-R BS.1116-3, “Methods for the subjective assessment of small impairments in audio systems”, 2015 Rec. ITU-R BS.1387-1, “Method for objective measurements of perceived audio quality”, 2001 Rec. ITU-R BS.2051, “Advanced sound system for programme production”, 2014 J.LIEBETRAU etc, “Standardization of PEAQ-MC: Extension of ITU-R BS.1387-1 to multichannel audio”, J. Audio Eng. Soc. 40th International Conference, 2010

However, the above-mentioned Non-Patent Document 4 describes the ITU-R recommendation BS of the above-mentioned Non-Patent Document 2 when objectively evaluating the quality of a multi-channel acoustic signal used in a multi-channel acoustic system having more than two channels. Unlike the objective evaluation method defined in 1387-1, a recognition model using the time difference between both ears and the level difference between both ears is used. Further, the objective evaluation result obtained by Non-Patent Document 4 is obtained from the above-mentioned ITU-R Recommendation BS of Non-Patent Document 1. It is not a value that sufficiently reflects the subjective evaluation result obtained by the subjective evaluation method specified in 1116-3. For this reason, ITU-R (International Telecommunication Union Radiocommunication Sector) has attempted standardization using the method of Non-Patent Document 4 described above, but it has not been approved and continues to the present.

By the way, in a multi-channel acoustic system having more than two channels, when subjectively evaluating an acoustic signal deteriorated by coding or the like, human beings are not good at concentrating and comparing acoustic signals in all directions. Therefore, when the number of channels of the acoustic signal is large, the subjective evaluation value tends to increase.

Also, with regard to content in which the sound image moves, humans are not good at concentrating and comparing acoustic signals in all directions while memorizing them with their heads. Therefore, when the number of channels is large, the subjective evaluation value tends to increase as well.

Since the multi-channel acoustic signal is presented to humans, it is desirable that the objective evaluation value is a value that reflects such a tendency of the subjective evaluation value. That is, it is desirable that the method for objectively evaluating the quality of a multi-channel acoustic signal having more than two channels is an objective evaluation method in consideration of the influence on the subjective evaluation value.

The above-mentioned ITU-R recommendation BS of Non-Patent Document 2. The objective evaluation method defined in 1387-1 is an objective evaluation method in consideration of the influence on the subjective evaluation value, but is a method applied to a two-channel acoustic signal and applied to a multi-channel acoustic signal exceeding two channels. Not the way.

Here, in addition to the method of the above-mentioned non-patent document 4, the ITU-R recommendation BS of the above-mentioned non-patent document 2 is used. A new method incorporating the objective evaluation method defined in 1387-1 can be envisioned. In this assumed method, the head impulse response HRIR is convolved with the original sound and the deteriorated sound of the multi-channel acoustic signal, and the convolution results of the original sound and the deteriorated sound are added to generate a 2-channel signal, and the 2-channel signal is used. , PEAQ The objective evaluation value is obtained by the objective sound quality measurement method.

This assumption method uses the PEAQ objective sound quality measurement method for objectively measuring the sound quality corresponding to the subjective evaluation. As shown in the experimental result of FIG. 10 described later, the objective evaluation result is the subjective evaluation result. The value is not close to.

As a factor that makes the subjective evaluation result different from the objective evaluation result, the deterioration of each channel is also added to the added acoustic signal, but the evaluator evaluates all of them with the same accuracy as the stereo signal. It is inferred that it is difficult.

Therefore, the present invention has been made to solve the above-mentioned problems, and an object thereof is a multi-channel capable of obtaining an objective evaluation result close to a subjective evaluation result with respect to the quality of a multi-channel acoustic signal having more than two channels. The purpose is to provide an objective evaluation device and a program.

In order to solve the above problems, the multi-channel objective evaluation device according to claim 1 is a multi-channel objective evaluation device that objectively evaluates a multi-channel acoustic signal having more than two channels, and each acoustic signal constituting the multi-channel acoustic signal is used. A convolutional signal output unit that outputs a head impulse response (HRIR) or a binoral chamber impulse response (BRIR) that represents the propagation characteristics of each channel as a convolution signal, and the original sound and deterioration of the multi-channel acoustic signal. Along with inputting the sound, the convolution signal for each channel output by the convolution signal output unit is input, the convolution signal is convoluted into the original sound for each channel, and all channels are convoluted based on the convolution results of all channels. In addition to generating a basic signal common to the above, the convolution signal is convoluted into the deterioration sound of one or a plurality of channels including the channel for each channel to generate a first convolution result, and the above 1 of all channels is generated. Alternatively, the convolution signal is convoluted into the original sound of a channel other than the plurality of channels to generate a second convolution result, and a signal to be measured is generated based on the first convolution result and the second convolution result. Each time, a signal processing unit that generates a binoral signal composed of the basic signal and the measured signal and the binoral signal for each channel generated by the signal processing unit are input, and for each channel, the binoral of the channel is input. Based on the signal, an evaluation unit that generates an objective evaluation result using a predetermined PEAQ (Perceptual Evaluation of Audio Quality) objective sound quality measurement method, and the objective evaluation result for each channel generated by the evaluation unit. It is characterized by including a multi-channel evaluation unit that generates an objective evaluation result of the multi-channel acoustic signal as a multi-channel objective evaluation result.

Further, in the multi-channel objective evaluation device according to claim 2, in the multi-channel objective evaluation device according to claim 1, the convolution signal output unit determines the number and arrangement of channels of the multi-channel acoustic signal. Is input, the convolution signal for each channel corresponding to the acoustic method is read out from a preset database and output, and the channel of the acoustic method and the convolution signal corresponding to the channel are stored in the database. It is characterized by being stored.

Further, in the multi-channel objective evaluation device according to claim 3, in the multi-channel objective evaluation device according to claim 1, the convolution signal output unit determines the reproduction position of each acoustic signal constituting the multi-channel acoustic signal. Information on the angle for each predetermined channel is input, the convolution signal for each channel corresponding to the angle for each channel is read out from a preset database, and the convolution signal is output to the database at the angle and the angle. It is characterized in that the corresponding convolution signal is stored.

Further, the multi-channel objective evaluation device according to claim 4 is the multi-channel objective evaluation device according to any one of claims 1 to 3, wherein the multi-channel evaluation unit is generated for each channel generated by the evaluation unit. It is characterized in that the lowest value among the objective evaluation results of the above is detected and the lowest value is generated as the multi-channel objective evaluation result.

Further, the multi-channel objective evaluation device according to claim 5 is the multi-channel objective evaluation device according to any one of claims 1 to 3, wherein the multi-channel evaluation unit is generated for each channel generated by the evaluation unit. The objective evaluation result is multiplied by a weighting coefficient for each predetermined channel, the multiplication result for each channel is added, and the addition result is generated as the multi-channel objective evaluation result.

The program of claim 6 is characterized in that the computer functions as the multi-channel objective evaluation device according to any one of claims 1 to 5.

As described above, according to the present invention, it is possible to obtain an objective evaluation result close to a subjective evaluation result for the quality of a multi-channel acoustic signal having more than two channels.

It is a block diagram which shows the structural example of the multi-channel objective evaluation apparatus by embodiment of this invention. It is a flowchart which shows the processing example of the multi-channel objective evaluation apparatus. It is a flowchart which shows the processing example of the convolution signal output part. It is a figure which shows the data structure example of DB. It is a flowchart which shows the 1st processing example of a signal processing part. It is a flowchart which shows the 2nd processing example of a signal processing part. It is a flowchart which shows the 1st processing example of a multi-channel evaluation part. It is a flowchart which shows the 2nd processing example of a multi-channel evaluation part. It is a flowchart which shows the setting processing example of the weighting coefficient W _{1 to 24} by a multi-channel evaluation unit. It is a figure which shows the experimental result.

Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the drawings.
[Outline of the invention]
When subjectively evaluating an acoustic signal (hereinafter referred to as "deteriorated sound") deteriorated by coding or the like, human beings tend to pay attention to the deterioration of the sound quality of each sound source. Further, in a multi-channel acoustic system, a sound source for reproducing a multi-channel acoustic signal is mixed at the highest level for a certain channel (for example, a frontal channel or a paired channel).

In view of such a situation, in the multi-channel objective evaluation device of the embodiment of the present invention, in order to approximate the degree of sound quality deterioration of a certain channel to the subjective evaluation, only a predetermined channel is used as the deteriorated sound and the other channels are used as the original sound. Treat as. Then, the multi-channel objective evaluation device generates a binaural signal using these deteriorated sounds and original sounds, and uses this binaural signal as an input signal for objective evaluation to perform objective evaluation.

Specifically, the multi-channel objective evaluation device inputs the original sound and the deteriorated sound of each acoustic signal constituting the multi-channel acoustic signal having more than two channels. Then, the multi-channel objective evaluation device performs a convolution process for each channel using, for example, all the original sounds and the deteriorated sound of only the channel, and generates a binaural signal for each channel in consideration of the subjective evaluation.

The multi-channel objective evaluation device uses a binoural signal as an input signal to be objectively evaluated, and uses the above-mentioned ITU-R recommendation BS of Non-Patent Document 2 for each channel. The objective evaluation value is obtained by the objective evaluation method specified in 1387-1. Then, the multi-channel objective evaluation device obtains a multi-channel objective evaluation value based on the objective evaluation value for each channel.

As a result, since the binaural signal to be objectively evaluated is a signal considering the subjective evaluation generated by paying attention to the deterioration of the sound quality of each sound source, the multi-channel objective evaluation value generated from the objective evaluation value of the binaural signal. Is a value close to the subjective evaluation value. Therefore, it is possible to obtain an objective evaluation result close to the subjective evaluation result for the quality of the multi-channel acoustic signal having more than two channels.

[Multi-channel objective evaluation device]
First, the configuration and processing of the multi-channel objective evaluation device according to the embodiment of the present invention will be described. FIG. 1 is a block diagram showing a configuration example of a multi-channel objective evaluation device according to an embodiment of the present invention.

This multi-channel objective evaluation device 1 is a device that objectively evaluates a multi-channel acoustic signal having more than two channels, and is described in the above-mentioned ITU-R recommendation BS of Non-Patent Document 2. Utilizing the objective evaluation method defined in 1387-1, the ITU-R recommendation BS of Non-Patent Document 1 described above. A multi-channel objective evaluation value z (multi-channel objective evaluation result) close to the subjective evaluation value obtained by the subjective evaluation method defined in 1116-3 is obtained. The multi-channel objective evaluation device 1 includes a convolution signal output unit 10, a signal processing unit 11, a PEAQ evaluation unit 12, and a multi-channel evaluation unit 13.

The multi-channel objective evaluation device 1 inputs the original sound x _{1 to 24} and the deteriorated sound x ' _{1 to 24} of the multi-channel acoustic signal, inputs the reproduction position information P, and based on the reproduction position information P, for each channel. Identify the convolution signal. Then, the multi-channel objective evaluation device 1 generates a binaural signal for each channel in consideration of subjective evaluation, evaluates the binaural signal in PEAQ, and based on the result, obtains a multi-channel objective evaluation value z in consideration of subjective evaluation. calculate.

Hereinafter, as an example of the multi-channel acoustic signal, an acoustic signal when the acoustic method is 22.2ch will be specifically described. The 22.2ch multi-channel acoustic signal is composed of 24 channels of acoustic signals.

The reproduction position information P is information regarding the reproduction position of each acoustic signal in the multi-channel acoustic system, and is, for example, information on the acoustic method of the multi-channel acoustic signal or information on an angle relating to the reproduction position. In the case of this example, 22.2ch acoustic method information is input as the reproduction position information P. The acoustic method uniquely determines the number and arrangement of channels. Alternatively, as the reproduction position information P, the angle consisting of the elevation angle (for each channel) and the azimuth angle (horizontal plane angle and vertical plane angle) for each acoustic signal constituting the 22.2ch multi-channel acoustic signal. Information is entered.

FIG. 2 is a flowchart showing a processing example of the multi-channel objective evaluation device 1. The multi-channel objective evaluation device 1 inputs the original sound x _{1 to 24} and the deteriorated sound x ' _{1 to 24} of each acoustic signal constituting the multi-channel acoustic signal, and also inputs the reproduction position information P (step S201). The multi-channel objective evaluation device 1 objectively evaluates the deteriorated sound x ' _{1 to 24} based on the original sound x _{1 to 24} of the multi-channel acoustic signal.

The multi-channel objective evaluation device 1 specifies, for example, head-related impulse responses HRIRs _{1 to 24} representing propagation characteristics for each channel as convolution signals for each channel based on the reproduction position information P (step S202).

The multi-channel objective evaluation device 1 performs convolution processing in consideration of subjective evaluation based on the original sound x _{1 to 24} for each channel, the deteriorated sound x ' _{1 to 24} , and the head impulse response HRIR _{1 to 24} , and for each channel. Binaural signals y _{1_ori to 24_ori} and y _{1_sig to 24_sig} are generated (step S203).

The multi-channel objective evaluation device 1 is based on the binaural signals y _{1_ori to 24_ori} and y _{1_sig to 24_sig} of the channel for each channel, and is based on the above-mentioned ITU-R recommendation BS of Non-Patent Document 2. PEAQ evaluation is performed by the objective evaluation method specified in 1387-1 (step S204). Then, the multi-channel objective evaluation device 1 obtains objective evaluation values z _{1 to 24} for each channel.

The multi-channel objective evaluation device 1 calculates and outputs a multi-channel objective evaluation value z based on the objective evaluation values z _{1 to 24} for each channel (step S205).

(Convolution signal output unit 10)
With reference to FIG. 1, the convolution signal output unit 10 includes a preset database (DB, not shown). The convolution signal output unit 10 inputs the reproduction position information P of the acoustic signal of 24 channels, and inputs the convolution signal for each channel corresponding to the reproduction position information P, for example, the head impulse response HRIR _{1 to 24} for each channel from the DB. read out. Then, the convolution signal output unit 10 outputs the head impulse responses HRIR _{1 to 24} for each channel to the signal processing unit 11.

FIG. 3 is a flowchart showing a processing example of the convolution signal output unit 10. The convolution signal output unit 10 inputs the reproduction position information P (step S301), and determines whether the reproduction position information P includes the acoustic method information or the angle information (step S302).

It is assumed that the reproduction position information P includes information on either one of the acoustic method and the angle. In the case of a standardized acoustic method such as 22.2ch, 11.1ch, 7.1ch, 5.1ch, etc., as in Non-Patent Document 3, the playback position is fixed, so a preset is registered. Keep it. In this case, the reproduction position information P includes information on the acoustic method for identifying 22.2ch and the like. On the other hand, when the fixed acoustic method is not used, the reproduction position information P includes angle information for specifying the reproduction position for each channel.

When the convolution signal output unit 10 determines in step S302 that the reproduction position information P includes the acoustic method information (step S302: acoustic method), the convolution signal output unit 10 corresponds to the acoustic method included in the reproduction position information P from the DB. The head impulse response HRIRs _{1 to 24} are read out (step S303).

On the other hand, when the convolution signal output unit 10 determines in step S302 that the reproduction position information P includes the angle information (step S302: angle), the head corresponding to the angle included in the reproduction position information P from the DB. Read out the part impulse response HRIRs _{1 to 24} (step S304).

The convolution signal output unit 10 shifts from step S303 or step S304 and outputs the head impulse responses HRIR _{1 to 24} for each channel to the signal processing unit 11 (step S305).

FIG. 4 is a diagram showing an example of DB data configuration. This DB is a transfer function between the acoustic system, channel number (label), elevation angle, azimuth angle, and head impulse response HRIR (speaker position and human ear position), which is a convolution signal corresponding to these information. It consists of data from the corresponding impulse response).

The acoustic system is 22.2ch, 11.1ch, 7.1ch, 5.1ch, etc., and the channel number is a number corresponding to each acoustic signal of the acoustic system. The elevation angle is the angle formed by the line between the speaker position and the position of the human ear with the horizontal plane, and the azimuth angle is the angle formed by the line between the speaker position and the position of the human ear with the vertical plane. Generally, the front direction is an elevation angle of 0 degrees and an azimuth angle of 0 degrees.

In the DB shown in FIG. 4, when the acoustic method is 22.2ch, the channel number 3 (label is FC (front center)), the elevation angle 0 °, the azimuth angle 0 °, and the head impulse corresponding to these information are displayed. Response HRIR ₃ etc. are stored. In addition, data of acoustic methods such as 5.1ch other than 22.2ch are also stored in the DB, and when the acoustic method is 5.1ch, the channel number 3 (label is C (center)) and the elevation angle is 0. °, azimuth angle 0 °, and head impulse response HRIR ₃ and the like corresponding to this information are stored.

When the reproduction position information P including the information of the acoustic method of 22.2ch is input, the convolution signal output unit 10 searches the DB of FIG. 4 using the acoustic method of 22.2ch as a key in step S303. Then, the convolution signal output unit 10 reads out the head impulse responses HRIRs _{1 to 24} corresponding to the channel numbers 1 to 24 of the 22.2ch acoustic method from the DB, respectively.

Thereby, the convolution signal output unit 10 can specify the head impulse response HRIR _{1 to 24} for each channel corresponding to the 22.2ch acoustic method without being conscious of the angle of each channel. In this case, the DB may store the acoustic method, the channel number (label), and the head impulse response HRIR corresponding to these information.

Further, when the convolution signal output unit 10 inputs the reproduction position information P including the elevation angle and azimuth information for each channel, in step S304, the convolution signal output unit 10 searches the DB of FIG. 4 using the elevation angle and azimuth angle for each channel as keys. .. Then, the convolution signal output unit 10 reads out the head impulse responses HRIR _{1 to 24} corresponding to the elevation angle and the azimuth angle for each channel from the DB, respectively.

As a result, the convolution signal output unit 10 has head-related impulse response HRIR _{1 to} each channel corresponding to the angle of each channel for a multi-channel system in which two or more speakers are arbitrarily arranged without preset speaker arrangement. ₂₄ can be identified. In this case, the DB may store the elevation angle, the azimuth angle, and the head impulse response HRIR corresponding to these information.

(Signal processing unit 11)
Returning to FIG. 1, the signal processing unit 11 inputs the original sound x _{1 to 24} and the deteriorated sound x ' _{1 to 24} of the multi-channel acoustic signal, and receives the head impulse response HRIR _{1 to 24} from the convolution signal output unit 10. input.

The signal processing unit 11 performs convolution processing based on the original sound x _{1 to 24} , the deteriorated sound x ' _{1 to 24} , and the head impulse response HRIR _{1 to 24} , and the binaural signal y _{1_ori to 24_ori} for each channel in consideration of subjective evaluation. ， Y _{Generates 1_sig to 24_sig} . Specifically, the signal processing unit 11 may use the signal processing unit 11 for each channel, for example, all the original sounds x _{1 to 24} , the deteriorated sound x'of only a predetermined channel including the channel, or the deteriorated sound x'of only the channel. The convolution process is performed based on the deteriorated sound x') of a plurality of channels including the channel and the head impulse response HRIR _{1 to 24} . The signal processing unit 11 outputs binaural signals y _{1_ori to 24_ori} and y _{1_sig to 24_sig} for each channel to the PEAQ evaluation unit 12.

Here, assuming that the number of channels of the multi-channel acoustic signal is M (M = 24 in this example), each channel (M) binaural signals y _{1_ori to M_ori} or y _{1_sig to M_sig} for each channel is generated. The deteriorated sound x'or the original sound x having the number of channels N (<M) is used. M is a positive integer larger than 2, and the number of channels N of the deteriorated sound x'or the original sound x is 1 or more and smaller than the number of channels M of the multi-channel acoustic signal (1 ≦ N <M).

The number of channels N of the deteriorated sound x'or the original sound x includes the channel (channel of channel number k) when the binoral signals y _{k_ori} and y _{k_sig} of the channel of channel number k (k = 1 to M) are generated. The number of channels of 1 or 2 or more. In addition to the channel with the channel number k, the channels adjacent to the channel may be included, or the interchannel correlation may be calculated and selected from the channels having a large normalization correlation coefficient. Here, assuming that the signal of the channel number k is f (t) and the signal of the adjacent channel is g (t), the normalized correlation function σ _fg is calculated by the following mathematical formula (1). σ _f and σ _g are standard deviations of the signals f (t) and g (t).

The binaural signals y _{1_ori to 24_ori} and y _{1_sig to 24_sig} are composed of the basic signals y _{1_ori to 24_ori} corresponding to the original sounds x ₁ to 24 and the measured signals y _{1_sig to 24_sig} corresponding to the deteriorated sounds x ' _{1 to 24} . .. The basic signals y _{1_ori to 24_ori} and the measured signals y _{1_sig to 24_sig} are generated by the signal processing unit 11 in the processing example shown in FIG. 5 or FIG. 6 to be described later.

FIG. 5 is a flowchart showing a first processing example of the signal processing unit 11. In this first processing example, convolution processing is performed for each channel based on all the original sounds x _{1 to 24} , the deteriorated sound x'of the channel only, and the head impulse response HRIR _{1 to 24} , and the binaural signal y _{_ori} ,. This is an example of generating y _{1_sig to 24_sig} . It is assumed that the number of channels M of the multi-channel acoustic signal is 24 and the number of channels of the deteriorated sound x'is N = 1.

The signal processing unit 11 inputs the original sound x _{1 to 24} and the deteriorated sound x ' _{1 to 24} of the multi-channel acoustic signal, and also inputs the head impulse response HRIR _{1 to 24} from the convolution signal output unit 10 (step S501). ..

The signal processing unit 11 performs convolution processing using all the original sounds x _{1 to 24} and the head impulse response HRIR _{1 to 24} , and generates a common basic signal y _{_ori} (step S502).

Specifically, as shown in the following formula (2), the signal processing unit 11 convolves the head impulse response HRIR 1 to 24 for each channel into the original sound x _{1 to 24} for each channel, respectively, and the signal processing unit 11 convolves the head impulse response HRIR _{1 to 24} for each channel. The convolution results are added, and the addition result is generated as a common basic signal y _{_ori} .

Here, if the basic signal of the channel number k is y _{k_ori} , the original sound of the channel number i is x _i , and the head impulse response of the channel number i is HRIR _i , the basic signal y _{k_ori} is the same as y _{_ori} . k and i are integers from 1 to 24, respectively, and * indicates a convolution operation.

The signal processing unit 11 has, for each channel, the original sound x of the number of channels 23 (= MN = 24-1), the deteriorated sound x'of the number of channels 1 (= N), and the head impulse response HRIR of all the channels. The convolution process is performed using _{1 to 24} to generate the measured signals y _{1_sig to 24_sig} for each channel (step S503).

Specifically, the signal processing unit 11 convolves the head impulse response HRIR into the original sound x of the number of channels 23 other than the channel (referred to as the channel number k) for each channel, and obtains the convolution result of the number of channels 23. Addition is performed, and the addition result of the original sound x of the number of channels 23 is obtained. Then, the signal processing unit 11 convolves the head impulse response HRIR with the deteriorated sound x'of the number of channels 1 in the channel, and obtains the convolution result of the deteriorated sound x'of the number of channels 1.

The signal processing unit 11 adds the convolution result of the deteriorated sound x'of the channel number 1 (channel number k) to the addition result of the original sound x of the channel number 23, and the addition result is measured by the channel. The signal y _{k_sig} is used, and the measured signals y _{1_sig to 24_sig} for each channel are generated.

The signal processing unit 11 convolves the head impulse response HRIRs _{1 to 24} into the original sounds x _{1 to 24} , adds the convolution results of all channels, and convolves the head impulse response HRIR into the original sounds x of the channel. By subtracting the latter convolution result from the former addition result, the addition result of the original sound x having the number of channels 23 may be obtained. Then, the signal processing unit 11 adds the convolution result of the deteriorated sound x'of the number of channels 1 to the addition result of the original sound x of the number of channels 23, and generates the measured signals y _{1_sig to 24_sig} for each channel. This corresponds to the calculation of the mathematical formula (3) described later.

Here, the measured signal of the channel number k is y _{k_sig} , the original sound of the channel numbers i and k is x _i and x _k , respectively, and the head impulse response of the channel numbers i and k is HRIR _i , HRIR _k and the channel number k, respectively. The measured signal y _{k_sig} is expressed by the following formula, _{where x'k} is the deteriorated sound of.

The formula (3) is a formula in which the number of channels N = 1 of the deteriorated sound x'is assumed to be subjectively evaluated by a human being paying attention to one channel. However, in reality, depending on the type of sound source, a human may focus on two or more channels for subjective evaluation. In this case, the measured signal y _{k_sig} is calculated when the number of channels N> 1 of the deteriorated sound x'. When the number of channels N> 1 of the deteriorated sound x', the second term on the right side of the equation (3) is convolved with respect to the original sound x for the number of channels N, and the respective calculation results are subtracted. Further, in the third term on the right side of the mathematical formula (3), a convolution calculation is performed on the deteriorated sound x'for the number of channels N, and the respective calculation results are added.

The signal processing unit 11 outputs the basic signal y _{_ori} generated in step S502 and the measured signals y _{1_sig to 24_sig} generated in step S503 to the PEAQ evaluation unit 12 (step S504).

In this way, the basic signal y _{_ori} is generated by the convolution process using the original sounds x _{1 to 24} of all channels. Further, the measured signals y _{1_sig to 24_sig} are generated for each channel by a convolution process using the original sound x of 23 channels other than the channel and the deteriorated sound x'of 1 channel of the channel.

That is, the binoral signals y _{k_ori} and y _{k_sig} of the predetermined channel (channel number k) are the basic signal y _{_ori} based on the original sound x _{1 to 24} of all channels and the deteriorated sound x ' _{1 to 24} of all channels. Of these, it is composed of the measured signal y _{k_sig} based on the deteriorated sound _x'k of the channel. Therefore, the measured signals y _{1_sig to 24_sig} are binaural signals in consideration of subjective evaluation that focuses on the deterioration of sound quality of individual sound sources in multi-channel acoustics.

FIG. 6 is a flowchart showing a second processing example of the signal processing unit 11. In this second processing example, convolution processing is performed for each channel based on all the deteriorated sounds x ' _{1 to 24} , the original sound x of only the channel, and the head impulse response HRIR _{1 to 24} , and the binaural signal y _{1_ori ~.} This is an example of generating _{24_ori} and y _{_sig} . It is assumed that the number of channels M of the multi-channel acoustic signal is 24 and the number of channels of the deteriorated sound x'is N = 1.

The signal processing unit 11 inputs the original sound x _{1 to 24} and the deteriorated sound x ' _{1 to 24} of the multi-channel acoustic signal, and also inputs the head impulse response HRIR _{1 to 24} from the convolution signal output unit 10 (step S601). ..

The signal processing unit 11 performs convolution processing using all the deteriorated sounds x ' _{1 to 24} and the head impulse response HRIR _{1 to 24} , and generates a common measured signal y _{_sig} (step S602).

Specifically, as shown in the following formula (4), the signal processing unit 11 convolves the head impulse response HRIR _{1 to 24} for each channel into the deterioration sound x ' _{1 to 24} for each channel, and all of them. The convolution results of the channels are added, and the addition result is generated as a common measured signal y _{_sig} .

Here, assuming that the measured signal of channel number k is y _{k_sig} , the degraded sound of channel number _i is x'i, and the head impulse response of channel number i is HRIR _i , the measured signal is y _{k_sig} and y _{_sig} . Will be the same.

The signal processing unit 11 performs convolution processing for each channel using the deteriorated sound x'of the number of channels 23, the original sound x of the number of channels 1, and the head impulse responses HRIR _{1 to 24} of all the channels, and performs the convolution processing for each channel. The basic signals y _{1_ori to 24_ori} are generated (step S603).

Specifically, the signal processing unit 11 convolves the head impulse response HRIR into the deteriorated sound x'of the number of channels 23 other than the channel, and adds the convolution result of the number of channels 23 to the number of channels 23. The addition result of the deteriorated sound x'is obtained. Then, the signal processing unit 11 convolves the head impulse response HRIR with the original sound x having the number of channels 1 in the channel, and obtains the convolution result of the original sound x having the number of channels 1.

The signal processing unit 11 adds the convolution result of the original sound x of the channel number 1 (channel number k) to the addition result of the deteriorated sound x'of the channel number 23, and the addition result is the basic signal of the channel. Let y _{k_ori} and generate basic signals y _{1_ori to 24_ori} for each channel.

The signal processing unit 11 convolves the head impulse response HRIRs _{1 to 24} with the deterioration sound x ' _{1 to 24} , adds the convolution results of all channels, and adds the head impulse response to the deterioration sound x'of the channel. By convolving the HRIR and subtracting the convolution result of the latter from the addition result of the former, the addition result of the deteriorated sound x'of the number of channels 23 may be obtained. Then, the signal processing unit 11 adds the convolution result of the original sound x of the number of channels 1 to the addition result of the deteriorated sound x'of the number of channels 23, and generates the basic signals y _{1_ori to 24_ori} for each channel. This corresponds to the calculation of the mathematical formula (5) described later.

Here, the basic signal of the channel number k is y _{k_ori} , the degraded sound of the channel numbers i and k is x'i and x'k, respectively, and the head impulse response of the channel numbers _{i and k is HRIR i , HRIR k and the} channel, respectively. Assuming that the original sound of the number k is x _k , the basic signal y _{k_ori} is expressed by the following formula.

When the number of channels of the deteriorated sound N> 1, the second term on the right side of the formula (5) is subjected to a convolution calculation for the deteriorated sound x'for the number of channels N, and the respective calculation results are subtracted. .. Further, in the third term on the right side of the mathematical formula (5), a convolution operation is performed on the original sound x for the number of channels N, and the respective calculation results are added.

The signal processing unit 11 outputs the measured signal y _{_sig} generated in step S602 and the basic signals y _{1_ori to 24_ori} generated in step S603 to the PEAQ evaluation unit 12 (step S604).

In this way, the measured signal y _{_sig} is generated by the convolution process using the deteriorated sounds x ' _{1 to 24} of all channels. Further, the basic signals y _{1_ori to 24_ori} are generated for each channel by a convolution process using the deteriorated sound x'of the number of channels 23 other than the channel and the original sound x of the channel number 1 of the channel.

That is, the binoral signals y _{k_ori} and y _{k_sig} of the predetermined channel (channel number k) are the measured signal y _{_sig} based on the deteriorated sound x ' _{1 to 24} of all channels and the original sound x _{1 to} all channels. Of the ₂₄ , it is composed of the basic signal y _{k_ori} based on the original sound x _k of the channel. In this case, the basic signal y _{k_ori} becomes the basic signal, and the measured signal y _{_sig} becomes a signal reflecting the deterioration of the sound quality of the sound source of the predetermined channel. Therefore, the basic signals y _{1_ori to 24_ori} are binaural signals in consideration of subjective evaluation that focuses on the deterioration of sound quality of individual sound sources.

Note that FIGS. 5 and 6 are examples of the number of channels N = 1 of the deteriorated sound x', but the same can be applied when N> 1. When N> 1, the signal processing unit 11 needs to select the original sound x having the number of channels N> 1 when generating the basic signal y _{k_ori} of the channel with the channel number k.

For the channel of channel number k, the signal processing unit 11 selects, for example, the original sound x of a predetermined number of channels adjacent to the channel in addition to the original sound x _k of the channel. The predetermined number is an integer of 1 or more.

Specifically, when there are a plurality of channels adjacent to the channel of channel number k, the signal processing unit 11 has a normalized correlation coefficient ρ _fg between the channel of channel number k and the channel adjacent thereto (the above-mentioned formula). (1)) is calculated for each of a plurality of adjacent channels. The signal processing unit 11 arranges a plurality of adjacent channels in descending order of the normalized correlation coefficient ρ _fg . The signal processing unit 11 selects the original sound x of a predetermined number of channels having a large normalization correlation coefficient ρ _fg in addition to the original sound x _k of the channel having the channel number k. It is assumed that a plurality of channels adjacent to the channel with the channel number k are preset from the reproduction position information P.

In this case, the signal processing unit 11 may select the original sound x of the channel not adjacent to the channel of the channel number k. Specifically, the signal processing unit 11 calculates the normalized correlation coefficient ρ _fg between the channel having the channel number k and the non-adjacent channel for channels other than the plurality of adjacent channels (non-adjacent channels). Then, when the normalized correlation coefficient ρ _fg is larger than that of the adjacent channel, the signal processing unit 11 selects the original sound x of the non-adjacent channel instead of the adjacent channel.

(PEAQ evaluation unit 12)
Returning to FIG. 1, the PEAQ evaluation unit 12 inputs binaural signals y _{1_ori to 24_ori} and y _{1_sig to 24_sig} for each channel from the signal processing unit 11. Then, the PEAQ evaluation unit 12 uses the above-mentioned ITU-R recommendation BS of Non-Patent Document 2 for each channel. The objective evaluation values z _{1 to 24} are obtained by the PEAQ objective sound quality measurement method, which is the objective evaluation method defined in 1387-1. The PEAQ evaluation unit 12 outputs the objective evaluation values z _{1 to 24} for each channel to the multi-channel evaluation unit 13.

The PEAQ evaluation unit 12 includes PEAQ evaluation means 20-1, PEAQ evaluation means 20-2, ..., And PEAQ evaluation means 20-24. The PEAQ evaluation means 20-k inputs the binoral signals y _{k_ori} and y _{k_sig} of the channel number k from the signal processing unit 11, obtains the objective evaluation value z _k using the algorithm of the PEAQ objective sound quality measurement method, and obtains the objective evaluation value z. _k is output to the multi-channel evaluation unit 13. k is an integer from 1 to 24 as described above.

Specifically, the PEAQ evaluation means 20-k inputs binaural signals y _{k_ori} and y _{k_sig} composed of the basic signal y _{k_ori} and the measured signal y _{k_sig} . Then, the PEAQ evaluation means 20-k generates an auditory model output signal for the basic signal y _{k_ori} and an auditory model output signal for the measured signal y _{k_sig} using an auditory model that reflects the perceptual characteristics of the human ear. do.

In this auditory model, the input signal is subjected to FFT (Fast Fourier Transform) by the algorithm that simulates the functions of the outer ear, middle ear, and inner ear to generate a frequency component signal, and the frequency component signal is generated. Hearing by classifying into groups that reflect the function of the inner ear, adding physiological noise such as blood flow to the signal of the frequency component, and calculating the nerve excitement pattern in consideration of the spread on the frequency axis and the time axis. Generate a model output signal.

The PEAQ evaluation means 20-k calculates the auditory distortion characteristics based on the respective auditory model output signals for the basic signal y _{k_ori} and the measured signal y _{k_sig} , and calculates the model output value indicating the degree of acoustic signal deterioration. Ask. Then, the PEAQ evaluation means 20- _k obtains an objective evaluation value zk based on the model output value by using a recognition model having a neural network structure.

A method of obtaining an objective evaluation value z _k using an algorithm of the PEAQ objective sound quality measurement method is known, and for details, refer to, for example, the above-mentioned Non-Patent Document 2 or the following documents.
Kaoru Watanabe, "Evaluation Method for Deterioration of Audio Signals", Journal of Acoustical Society of Japan, Vol. 63, No. 11 (2007), pp.686-692

(Multi-channel evaluation unit 13)
The multi-channel evaluation unit 13 inputs objective evaluation values z _{1 to 24} for each channel from the PEAQ evaluation unit 12, obtains multi-channel objective evaluation values z based on objective evaluation values z _{1 to 24} , and multi-channel objective. The evaluation value z is output.

FIG. 7 is a flowchart showing a first processing example of the multi-channel evaluation unit 13. The first processing example is an example in which the lowest value z _L among the objective evaluation values z _{1 to 24} for each channel obtained by the PEAQ objective sound quality measurement method is set as the multi-channel objective evaluation value z.

The multi-channel evaluation unit 13 inputs objective evaluation values z _{1 to 24} for each channel from the PEAQ evaluation unit 12 (step S701), and detects the lowest value z _L among the objective evaluation values z _{1 to 24} for each channel. (Step S702).

The multi-channel evaluation unit 13 sets the lowest value z _L detected in step S702 to the multi-channel objective evaluation value z (z = z _L ), and outputs the multi-channel objective evaluation value z (step S703).

In this way, the multi-channel evaluation unit 13 outputs the lowest value z _L among the objective evaluation values z _{1 to 24} for each channel obtained by the PEAQ objective sound quality measurement method as the multi-channel objective evaluation value z. did. As a result, the objective evaluation value of the channel having the lowest evaluation when a human pays attention to a specific channel in the multi-channel sound is output as the multi-channel objective evaluation value z. That is, the multi-channel objective evaluation value z is close to the subjective evaluation value evaluated by paying attention to the deterioration of the sound quality of each sound source.

FIG. 8 is a flowchart showing a second processing example of the multi-channel evaluation unit 13. In the second processing example, the objective evaluation values z _{1 to 24} for each channel obtained by the PEAQ objective sound quality measurement method are multiplied by the weighting coefficients W _{1 to 24} , and the multiplication results of all channels are added to obtain a multi-channel. This is an example of obtaining the objective evaluation value z.

The multi-channel evaluation unit 13 inputs objective evaluation values z _{1 to 24} for each channel from the PEAQ evaluation unit 12 (step S801), and a predetermined weighting coefficient W _{1 to 24} is set in the objective evaluation values z _{1 to 24} for each channel. Are multiplied by each, and the multiplication result for each channel is obtained (step S802). The total value of the weighting coefficients W1 _{to 24} is 1.

The multi-channel evaluation unit 13 adds the multiplication results of all the channels obtained in step S802 (step S803), sets the addition result to the multi-channel objective evaluation value z, and outputs the multi-channel objective evaluation value z. (Step S804).

Here, the second processing example shown in FIG. 8 is represented by the following mathematical formula.

As the predetermined weighting coefficients W _{1 to 24} , a smaller value is used as the objective evaluation value z _{1 to 24} is larger (the smaller the deterioration is), and the smaller the objective evaluation value z 1 to 24 is (the deterioration is less) for each channel. Larger values are used. The predetermined weighting coefficients W1 _{to 24} may be preset by the user or may be automatically set by a predetermined process.

Hereinafter, an example of setting the weighting coefficients W1 _{to 24} in a predetermined process will be described. FIG. 9 is a flowchart showing an example of setting processing of the weighting coefficients W1 _{to 24} by the multi-channel evaluation unit 13. The multi-channel evaluation unit 13 sets channel numbers i (i = 1 to 24) in order (step S901), and determines whether or not the objective evaluation value z _i is larger than a predetermined value (step S902).

In the objective evaluation value z _i obtained by the PEAQ evaluation unit 12, 0 is "cannot detect deteriorated sound", -1 is "can detect deteriorated sound but does not bother", and -2 is "slightly worried about deteriorated sound". , -3 indicates "I am concerned about the deteriorated sound", and -4 indicates "I am very concerned about the deteriorated sound", the predetermined value used in step S902 is, for example, -1.

When the multi-channel evaluation unit 13 determines in step S902 that the objective evaluation value _zi is larger than the predetermined value (step S902: Y), the multi-channel evaluation unit 13 determines that the acoustic signal of the channel of the channel number i has little deterioration. The weighting coefficient W _i = 0 is set (step S903).

On the other hand, when the multi-channel evaluation unit 13 determines in step S902 that the objective evaluation value z _i is not larger than the predetermined value (step S902: N), the weighting coefficient W _i is based on the loudness level of the acoustic signal. Is set (step S904).

Specifically, the multi-channel evaluation unit 13 inputs the loudness level for the acoustic signal of the channel of channel number i from the loudness measurement unit (not shown in FIG. 1). When there are a plurality of channels that the multi-channel evaluation unit 13 determines that the loudness level is not larger than the predetermined value, the larger the loudness level, the larger the weighting coefficient Wi _i (closer to 1), and the smaller the loudness level. The weighting coefficient W _i is set based on the loudness level of the acoustic signal so that the weighting coefficient W _i becomes smaller (closer to 0). As a result, a weighting coefficient Wi corresponding to the _loudness level of the acoustic signal can be obtained. It is assumed that the total value of the weighting coefficients W _{1 to 24} is 1.

In this case, the loudness measuring unit (not shown) measures loudness (loudness) for each channel by using, for example, the method of the following literature.
Rec. ITU-R BS.1770-4, “Algorithms to measure audio programme loudness and true-peak audio level”
The loudness measuring unit outputs the loudness level for each of a plurality of channels to the multi-channel evaluation unit 13. The multi-channel evaluation unit 13 sets a weighting coefficient Wi _i according to the loudness level for each channel.

The multi-channel evaluation unit 13 may set a _weighting coefficient Wi corresponding to the channel normalization correlation coefficient ρ _fg . Specifically, the multi-channel evaluation unit 13 inputs the normalized correlation coefficient ρ _fg in the channel of channel number i from the correlation coefficient calculation unit (not shown in FIG. 1). In the multi-channel evaluation unit 13, the weighting coefficient Wii becomes larger (closer to 1) as the normalization correlation coefficient ρ _fg becomes larger, and the weighting coefficient _{Wi i becomes} smaller as the normalization correlation coefficient ρ _fg becomes smaller. (To be close to 0), set the weighting factor Wi _i . As a result, a weighting coefficient Wi corresponding to the channel normalization correlation coefficient _{ρ fg} can be obtained.

In this case, the correlation coefficient calculation unit (not shown) calculates the normalized correlation coefficient ρ _fg between the channel of channel number i and the channel other than the channel, respectively, using the above formula (1). Then, the correlation coefficient calculation unit outputs this to the multi-channel evaluation unit 13 as a normalized correlation coefficient ρ _fg in the channel of channel number i.

Further, the multi-channel evaluation unit 13 detects the lowest value z _L among the objective evaluation values z _i for each channel, and the objective evaluation value z _i is predetermined for a plurality of channels adjacent to the channel having the lowest value z _l . If it is less than or equal to the value, the weighting coefficient W _i may be set so that the total value of the weighting coefficient W _i exceeds 1. However, the total value of the weighting coefficient W _i shall not exceed 2. Further, when the objective evaluation value z _i obtained by the PEAQ evaluation unit 12 is represented by 0 to -4 as described above, the predetermined value to be compared with the objective evaluation value z _i is, for example, -1.

In this way, the multi-channel evaluation unit 13 multiplies the objective evaluation values z _{1 to 24} for each channel obtained by the PEAQ objective sound quality measurement method by the predetermined weighting coefficients W _{1 to 24} , and adds all the multiplication results. Therefore, the multi-channel objective evaluation value z is generated and output. As a result, for the objective evaluation values z _{1 to 24} for each channel obtained by the PEAQ objective sound quality measurement method when a human focuses on a specific channel, weighting coefficients W _{1 to 24} according to the degree of attention are used. As a result, a multi-channel objective evaluation value z that reflects the degree of attention that differs for each channel is generated and output. That is, the multi-channel objective evaluation value z is close to the subjective evaluation value evaluated by paying attention to the deterioration of the sound quality of each sound source.

〔Experimental result〕
Next, the experimental results by computer simulation will be described. In this experimental result, the multi-channel objective evaluation value z output by the multi-channel objective evaluation device 1 is the ITU-R recommendation BS of Non-Patent Document 1 described above. It shows that it is close to the subjective evaluation value obtained by the subjective evaluation method defined in 1116-3.

FIG. 10 is a diagram showing the experimental results, and shows the results of evaluating the environmental sound of the 22.2ch multi-channel acoustic signal actually picked up. (A) is the above-mentioned ITU-R recommendation BS of Non-Patent Document 1. The subjective evaluation result obtained by the subjective evaluation method defined in 1116-3 is shown, and (b) shows the objective evaluation result (when the number of channels N = 1 of the deteriorated sound x') according to the embodiment of the present invention.

Further, (c) is based on the method of the above-mentioned non-patent document 4 and the above-mentioned ITU-R recommendation BS of the above-mentioned non-patent document 2. The objective evaluation result by the prior art (the above-mentioned assumed method) incorporating the objective evaluation method defined in 1387-1 is shown. Specifically, as described above, the objective evaluation result of (c) is obtained by convolving the head impulse response HRIR into the multi-channel acoustic signal to generate a 2-channel signal, and using the objective evaluation method of Non-Patent Document 2 described above. This is the result of the request.

(A) The horizontal axes of (b) and (c) indicate the bit rate [kbit / s] of the acoustic signal. The higher the bit rate, the lower the compression rate, and the lower the bit rate, the higher the compression rate. The vertical axis of (a) shows the subjective evaluation value (Diff Grade), and the vertical axis of (b) and (c) shows the objective evaluation value (Diff Grade). The objective evaluation value (b) is the multi-channel objective evaluation value z output by the multi-channel evaluation unit 13 of the multi-channel objective evaluation device 1 shown in FIG.

Similar to the above, 0 of the subjective evaluation value and the objective evaluation value is "cannot detect the deteriorated sound", -1 is "the deterioration sound can be detected but does not bother me", and -2 is "the deteriorated sound is a little worrisome". , -3 indicates "I'm worried about the deteriorated sound", and -4 indicates "I'm very worried about the deteriorated sound".

(A) From (b) and (c), the objective evaluation result of the embodiment of the present invention shown in (b) is the subjective evaluation result shown in (a) rather than the objective evaluation result of the prior art shown in (c). It turns out that it is close to.

As described above, by using the multi-channel objective evaluation device 1 of the embodiment of the present invention, the above-mentioned ITU-R recommendation BS of Non-Patent Document 1 can be used. It is possible to obtain a multi-channel objective evaluation value z close to the subjective evaluation value obtained by the subjective evaluation method defined in 1116-3.

As described above, according to the multi-channel objective evaluation device 1 of the embodiment of the present invention, the convolution signal output unit 10 uses a preset DB and is based on the reproduction position information P of the acoustic signal of 24 channels. The head impulse response HRIR _{1 to 24} for each channel is specified and output.

The signal processing unit 11 performs convolution processing based on the original sound x _{1 to 24} of the multi-channel acoustic signal, the deteriorated sound x ' _{1 to 24} , and the head impulse response HRIR _{1 to 24} , and the binaural for each channel in consideration of subjective evaluation. Generates signals y _{1_ori to 24_ori} and y _{1_sig to 24_sig} . Specifically, the signal processing unit 11 performs convolution processing for each channel, for example, based on all the original sounds x _{1 to 24} , the deteriorated sound x'only for the channel, and the head impulse response HRIR _{1 to 24} . Binaural signals y _{1_ori to 24_ori} and y _{1_sig to 24_sig} are generated.

The PEAQ evaluation unit 12 has the ITU-R recommendation BS of Non-Patent Document 2 described above based on the binaural signals y _{1_ori to 24_ori} and y _{1_sig to 24_sig} for each channel. The objective evaluation values z _{1 to 24} are obtained by the PEAQ objective sound quality measurement method, which is the objective evaluation method defined in 1387-1.

The multi-channel evaluation unit 13 obtains the multi-channel objective evaluation value z based on the objective evaluation values z _{1 to 24} for each channel.

Here, the binaural signals y _{1_ori to 24_ori} and y _{1_sig to 24_sig} used by the PEAQ evaluation unit 12 for objective evaluation are signals considering the subjective evaluation generated by the signal processing unit 11 focusing on the deterioration of the sound quality of individual sound sources. Is. As a result, the multi-channel objective evaluation value z obtained by the multi-channel evaluation unit 13 is generated from the objective evaluation values z _{1 to 24} of the binaural signals y _{1_ori to 24_ori} and y _{1_sig to 24_sig} , and is therefore close to the subjective evaluation value. It becomes. Therefore, it is possible to obtain an objective evaluation result close to the subjective evaluation result for the quality of the multi-channel acoustic signal having more than two channels.

Although the present invention has been described above with reference to embodiments, the present invention is not limited to the above-described embodiment and can be variously modified without departing from the technical idea. In the above-described embodiment, the multi-channel objective evaluation device 1 obtains the objective evaluation value z of the round channel with the 22.2ch multi-channel acoustic signal as the evaluation target. The present invention is not limited to the evaluation target of 22.2ch multi-channel acoustic signals, but is also applicable to multi-channel acoustic signals of other acoustic methods such as 11.1ch, 7.1ch, and 5.1ch.

Further, the present invention includes not only a multi-channel acoustic signal of an acoustic method such as 22.2ch in which the speaker arrangement is preset, but also a multi-channel acoustic signal in which two or more speakers in which the speaker arrangement is not preset are arbitrarily arranged. Also applies to.

Further, in the above embodiment, the multi-channel objective evaluation device 1 uses the head impulse response HRIR _{1 to 24} as the convolution signal. The present invention does not limit the convolution signal to the head impulse responses HRIRs _{1 to 24} , but other impulse responses such as binaural room impulse responses (BRIRs) _{1 to 24} may be used.

In this case, referring to FIG. 4, the DB provided in the convolution signal output unit 10 stores the binaural chamber impulse responses BRIRs _{1 to 24} instead of the head impulse responses HRIRs _{1 to 24} . The convolution signal output unit 10 reads out the binaural chamber impulse responses BRIRs _{1 to 24} representing the propagation characteristics of each channel corresponding to the reproduction position information P from the DB. Then, the signal processing unit 11 performs convolution processing based on the original sound x _{1 to 24} and the deteriorated sound x ' _{1 to 24} of the multi-channel acoustic signal, and the binaural room impulse response BRIR _{1 to 24} , and the channel in consideration of the subjective evaluation. Generates binaural signals y _{1_ori to 24_ori} and y _{1_sig to 24_sig} for each.

Further, in the above-described embodiment, the multi-channel objective evaluation device 1 convolves the binaural chamber impulse responses BRIR _{1 to 24} with the original sound x _{1 to 24} and the deteriorated sound x ' _{1 to 24} of the multi-channel acoustic signal, and the binaural signal for each channel. Changed to generate y _{1_ori to 24_ori} and y _{1_sig to 24_sig} . The present invention does not limit this convolution process to the calculation in the time domain, but the original sound x _{1 to 24} and the deteriorated sound x ' _{1 to 24} converted into the frequency domain, and the head related transfer function (HRTF). ) The product of _{1 to 24} may be calculated and converted into the time domain to generate the binoral signals y _{1_ori to 24_ori} and y _{1_sig to 24_sig} . Further, in the present invention, the product of the frequency components of the original sound x _{1 to 24} and the deteriorated sound x ' _{1 to 24} and the binaural room transfer function BRTF (Binaural Room Transfer Function) _{1 to 24} is calculated and converted into a time domain to be converted into a binaural. The signals y _{1_ori to 24_ori} and y _{1_sig to 24_sig} may be generated.

As the hardware configuration of the multi-channel objective evaluation device 1 according to the embodiment of the present invention, a normal computer can be used. The multi-channel objective evaluation device 1 is composed of a computer provided with a volatile storage medium such as a CPU and RAM, a non-volatile storage medium such as a ROM, and an interface.

Each function of the convolution signal output unit 10, the signal processing unit 11, the PEAQ evaluation unit 12, and the multi-channel evaluation unit 13 provided in the multi-channel objective evaluation device 1 is performed by causing the CPU to execute a program describing these functions. It will be realized.

These programs are stored in the storage medium, read by the CPU, and executed. In addition, these programs can be stored and distributed in storage media such as magnetic disks (floppy (registered trademark) disks, hard disks, etc.), optical disks (CD-ROM, DVD, etc.), semiconductor memories, etc., and can be distributed via a network. You can also send and receive.

1 Multi-channel objective evaluation device 10 Folded signal output unit 11 Signal processing unit 12 PEAQ evaluation unit 13 Multi-channel evaluation unit 20-1 to 24 PEAQ evaluation means x _{1 to} 24 Original sound of multi-channel acoustic signal x ' _{1 to 24} Multi-channel sound Signal deterioration sound P Playback position information HRIR _{1 to 24} Head impulse response M Number of channels of multi-channel acoustic signal N Number of channels of deterioration sound x'N
BRIR _{1 to 24} binaural chamber impulse response HRTF _{1 to 24} head related transfer function BRTF _{1 to 24} binaural chamber transfer function y _{1_ori to 24_ori} basic signal (binaural signal)
y _{1_sig ~ 24_sig} Measured signal (binaural signal)
y _{_ori} Common basic signal y _{_sig} Common measured signal z Objective evaluation value for each channel _{1 to 24} z Objective evaluation value for multi-channel ρ _fg Normalization correlation coefficient W _{1 to 24} Weighting coefficient

Claims (6)

In a multi-channel objective evaluation device that objectively evaluates multi-channel acoustic signals exceeding two channels
A convolution signal output that outputs a head-related impulse response (HRIR) or binaural chamber impulse response (BRIR) representing the propagation characteristics of each channel as a convolution signal corresponding to each channel of the acoustic signal constituting the multi-channel acoustic signal. Department and
The original sound and the deteriorated sound of the multi-channel acoustic signal are input, and the convolution signal for each channel output by the convolution signal output unit is input.
The convolution signal is convoluted into the original sound for each channel, and a basic signal common to all channels is generated based on the convolution results of all channels.
For each channel, the convolution signal is convoluted into the degraded sound of one or more channels including the channel to generate a first convolution result, and the original sound of a channel other than the one or the plurality of channels among all channels. The convolution signal is convoluted to generate a second convolution result, and a signal to be measured is generated based on the first convolution result and the second convolution result.
A signal processing unit that generates a binaural signal composed of the basic signal and the measured signal for each channel.
The binoral signal for each channel generated by the signal processing unit is input, and each channel is objectively measured using a predetermined PEAQ (Perceptual Evaluation of Audio Quality) objective sound quality measurement method based on the binoral signal of the channel. An evaluation unit that generates evaluation results and
A multi-channel evaluation unit that generates an objective evaluation result of the multi-channel acoustic signal as a multi-channel objective evaluation result based on the objective evaluation result for each channel generated by the evaluation unit.
A multi-channel objective evaluation device characterized by being equipped with. In the multi-channel objective evaluation device according to claim 1,
The convolution signal output unit is
Information on the acoustic method that determines the number and arrangement of channels of the multi-channel acoustic signal is input, and the convolution signal for each channel corresponding to the acoustic method is read out from a preset database and output.
The database is a multi-channel objective evaluation device, characterized in that a channel of the acoustic system and the convolution signal corresponding to the channel are stored. In the multi-channel objective evaluation device according to claim 1,
The convolution signal output unit is
The information of the angle for each channel that determines the reproduction position for each acoustic signal constituting the multi-channel acoustic signal is input, and the convolution signal for each channel corresponding to the angle for each channel is input from a preset database. Read and output,
A multi-channel objective evaluation device, characterized in that the database stores the angle and the convolution signal corresponding to the angle. In the multi-channel objective evaluation device according to any one of claims 1 to 3.
The multi-channel evaluation unit
A multi-channel objective evaluation device, characterized in that the lowest value among the objective evaluation results for each channel generated by the evaluation unit is detected and the lowest value is generated as the multi-channel objective evaluation result. In the multi-channel objective evaluation device according to any one of claims 1 to 3.
The multi-channel evaluation unit
The objective evaluation result for each channel generated by the evaluation unit is multiplied by a weighting coefficient for each predetermined channel, the multiplication result for each channel is added, and the addition result is generated as the multi-channel objective evaluation result. , A multi-channel objective evaluation device characterized by that. A program for making a computer function as the multi-channel objective evaluation device according to any one of claims 1 to 5.

Images