JP4886907B2 - Audio signal correction apparatus and audio signal correction method - Google Patents

Description

  The present invention relates to an audio signal correction technique for adaptively performing sound quality correction processing on an audio signal and a music signal included in an audio signal.

  As is well known, for example, in a broadcast receiving device that receives a television broadcast or an information reproducing device that reproduces recorded information from an information recording medium, the received broadcast signal or the signal read from the information recording medium When reproducing an audio signal, the audio signal is subjected to a sound quality correction process to further improve the sound quality.

  In this case, the content of the sound quality correction processing applied to the audio signal differs depending on whether the audio signal is a sound signal such as a human voice or a music (non-speech) signal such as a music piece. In other words, sound quality is improved by performing sound quality correction processing to emphasize and clarify the center localization component, such as talk scenes and sports conditions, for audio signals, and stereo for music signals. The sound quality is improved by applying a sound quality correction process with a feeling of emphasis.

  For this reason, it is considered to determine whether the acquired audio signal is a voice signal or a music signal, and perform a corresponding sound quality correction process according to the determination result. However, since an audio signal and a music signal are often mixed in an actual audio signal, it is difficult to discriminate between them, so that an appropriate sound quality correction process is performed on the audio signal. The current situation is not to say.

  Patent Document 1 determines whether an audio signal is speech or non-speech according to the degree of speech and music, and further determines whether the audio signal is speech or non-speech depending on whether the audio signal is a monaural signal or a stereo signal. A configuration for optimizing the determination is disclosed.

JP 2007-67858 A

  However, in the configuration of Patent Document 1, it is difficult to properly determine the signal content when the audio signal is a dual monaural signal or when the audio signal is a monaural transmission even if it is a stereo signal.

  An object of the present invention is to provide an audio signal correction apparatus and an audio signal correction apparatus that evaluate the contents of an input audio signal and perform adaptive sound quality correction processing.

  An audio signal correction apparatus according to an embodiment of the present invention determines whether an input audio signal is a monaural signal or a stereo signal based on channel information, and determines whether the input audio signal is an audio signal or a music signal. A feature extracting means for extracting a plurality of feature quantity parameters, and a voice indicating whether the input audio signal is close to a speech signal or a music signal based on the plurality of feature quantity parameters extracted by the feature extraction means A signal type determining means for calculating a music identification score, a level calculating means for calculating a sound level of the input audio signal and an output level of the music level using the sound music identification score, and the level calculated by the level calculating means Based on the output level, a sound quality correction process is performed on the input audio signal. With the door.

  ADVANTAGE OF THE INVENTION According to this invention, the audio signal correction apparatus which evaluates the content of an input audio signal and performs an adaptive sound quality correction process can be provided.

It is a block diagram which shows schematic structure of the digital television broadcast receiver which concerns on embodiment of this invention. It is a block diagram which shows schematic structure of the audio processing module which concerns on embodiment of this invention. It is a flowchart explaining the feature-value extraction process which concerns on embodiment of this invention. It is a flowchart explaining the signal type determination process which concerns on embodiment of this invention. It is a flowchart explaining the level calculation process which concerns on embodiment of this invention.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 shows a main signal processing system of the digital television broadcast receiver 11. That is, the satellite digital television broadcast signal received by the BS / CS (broadcasting satellite / communication satellite) digital broadcast receiving antenna 43 is supplied to the satellite digital broadcast tuner 45 via the input terminal 44. A broadcast signal of a desired channel is selected.

  The broadcast signal selected by the tuner 45 is demodulated into a digital video signal and an audio signal by being sequentially supplied to a PSK (phase shift keying) demodulation module 46 and a TS (transport stream) decoding module 47. And then output to the signal processing module 48.

  The terrestrial digital television broadcast signal received by the terrestrial broadcast receiving antenna 49 is supplied to the terrestrial digital broadcast tuner 51 via the input terminal 50, so that the broadcast signal of the desired channel is selected. Is done.

  The broadcast signal selected by the tuner 51 is demodulated into a digital video signal and an audio signal by being sequentially supplied to an OFDM (orthogonal frequency division multiplexing) demodulation module 52 and a TS decoding module 53 in, for example, Japan. After that, it is output to the signal processing module 48.

  The terrestrial analog television broadcast signal received by the terrestrial broadcast receiving antenna 49 is supplied to the terrestrial analog broadcast tuner 54 via the input terminal 50, so that the broadcast signal of the desired channel is selected. Bureau. The broadcast signal selected by the tuner 54 is supplied to the analog demodulation module 55, demodulated into an analog video signal and audio signal, and then output to the signal processing module 48.

  Here, the signal processing module 48 selectively performs predetermined digital signal processing on the digital video signal and audio signal supplied from the TS decoding modules 47 and 53, respectively, and the graphic processing module 56 and audio processing. Output to module 57.

  The signal processing module 48 is connected to a plurality (four in the illustrated case) of input terminals 58a, 58b, 58c, and 58d. These input terminals 58a to 58d can input analog video signals and audio signals from the outside of the digital television broadcast receiving apparatus 11, respectively.

  The signal processing module 48 selectively digitizes the analog video signal and audio signal supplied from the analog demodulation module 55 and the input terminals 58a to 58d, respectively, and performs the digitization on the digitized video signal and audio signal. After performing predetermined digital signal processing, the digital signal is output to the graphic processing module 56 and the audio processing module 57.

  The graphic processing module 56 has a function of superimposing and outputting the OSD signal generated by the OSD (on screen display) signal generation module 59 on the digital video signal supplied from the signal processing module 48. The graphic processing module 56 selectively outputs the output video signal of the signal processing module 48 and the output OSD signal of the OSD signal generation module 59, and combines both outputs so as to constitute half of the screen. Can be output.

  The digital video signal output from the graphic processing module 56 is supplied to the video processing module 60. The video processing module 60 converts the input digital video signal into an analog video signal in a format that can be displayed on the video display 14 and then outputs the analog video signal to the video display 14 to display the video. Derived outside through 61.

  The audio processing module 57 performs a sound quality correction process, which will be described later, on the input digital audio signal, and then converts it into an analog audio signal in a format that can be reproduced by the speaker 15. The analog audio signal is output to the speaker 15 for audio reproduction, and is derived to the outside via the output terminal 62.

  Here, in the digital television broadcast receiving apparatus 11, all operations including the above-described various receiving operations are comprehensively controlled by the control module 63. The control module 63 includes a CPU (central processing unit) 64 and receives operation information from the operation module 16 or operation information transmitted from the remote controller 17 and received by the light receiving module 18. Each unit is controlled to reflect the operation content.

  In this case, the control module 63 mainly includes a ROM (read only memory) 65 storing a control program executed by the CPU 64, a RAM (random access memory) 66 providing a work area to the CPU 64, and various setting information. And a non-volatile memory 67 in which control information and the like are stored.

  FIG. 2 shows a configuration in which a signal characteristic analysis module 70 and a sound quality correction module 80 are provided in the audio processing module 57. The signal characteristic analysis module 70 includes a feature extraction module 72, a signal type determination module 74, and a level calculation module 76. Furthermore, the feature extraction module 72 includes a first feature extraction module 72a and a second feature extraction module 72b. The signal type determination module 74 includes a first signal type determination module 74a and a second signal type determination module 74b. An input audio signal is supplied to the input terminal 71. The control module 63 supplies the input audio signal to the feature extraction module 72. The control module 63 supplies channel information (monaural / stereo signal information) of the input audio signal to each module constituting the signal characteristic analysis module 70.

  When the input audio signal is a stereo signal, the first feature extraction module 72a calculates various feature amount parameters for determining whether the input audio signal is a voice signal or a music signal. When the input audio signal is a monaural signal, the second feature extraction module 72b calculates various feature amount parameters for determining whether the input audio signal is a voice signal or a music signal. The feature extraction module 72 switches to the first feature extraction module 72a or the second feature extraction module 72b depending on whether the input audio signal is a stereo signal or a monaural signal.

  The first signal type determination module 74a determines whether the input audio signal (stereo signal) is an audio signal or a music signal. Similarly, the second feature extraction module 74b determines whether the input audio signal (monaural signal) is an audio signal or a music signal. The signal type determination module 74 switches to the first signal type determination module 74a or the second signal type determination module 74b depending on whether the input audio signal is a stereo signal or a monaural signal.

  The level calculation module 76 calculates a voice / music level including accuracy information for finely controlling the sound quality with respect to the voice signal or the music signal. The level calculation module 76 outputs the sound / music level information to the sound quality correction module 80.

  In the present embodiment, the first feature extraction module 72a and the second feature extraction module 72b are different modules, and the first signal type determination module 74a and the second feature extraction module 74b are different modules. Each may be integrated.

  The sound quality correction module 80 performs sound quality correction processing based on the music / sound level information calculated by the signal characteristic analysis module 70. The sound quality correction module 80 supplies the output audio signal subjected to the sound quality correction process to the output terminal 77.

  That is, the signal characteristic analysis module 70 and the sound quality correction module 80 identify a music section and a voice section without processing delay in receiving a broadcast or reproducing a content from a recording medium, and appropriately apply to an input audio signal according to the contents of the scene. It has a function of executing scene adaptive sound quality correction for improving the sound quality by performing a sound quality correction process.

  Next, operations of the first feature extraction module 72 and the second feature extraction module 73 will be described. FIG. 3 is a flowchart for explaining the feature amount extraction processing. First, the feature extraction module 72 cuts out an input audio signal as a frame every several hundreds of milliseconds. Furthermore, the feature extraction module 72 divides them into subframes every several tens of milliseconds (step S101). For example, one subframe is 20 msec.

  The feature extraction module 72 determines whether the number of channels of the input audio signal is 2 (that is, whether it is a monaural signal or a stereo signal) based on the channel information of the input audio signal (step S102). Here, for example, when the input audio signal demodulated from the broadcast signal selected by the tuner 51 is a multi-channel stereo signal, the signal processing module 48 performs a downmix process from the multi-channel to the 2-channel stereo signal. It is assumed that The signal processing module 48 supplies a two-channel stereo signal to the input terminal 71 as an input audio signal.

  When the number of channels is 2 (step S102, YES), the feature extraction module 72 determines whether or not the input audio signal is a normal stereo signal that is not a dual monaural signal (step S103). In the dual monaural signal, even if the number of channels of the dual monaural signal is 2, the sound superimposed on the main / sub channels is originally a separate monaural signal.

  When the input audio signal is a normal stereo signal that is not a dual monaural signal (step S103, YES), the feature extraction module 72 determines the power ratio (LR) of the left and right (LR) signals of 2-channel stereo in the input audio signal in units of subframes. Power ratio). Here, even if the format of the input audio signal is a stereo signal, there are cases where it is actually transmitted like a monaural signal. In this case, the LR channels have substantially the same signal, and the feature extraction module 72 cannot determine only by the number of channels. Therefore, the feature extraction module 72 calculates an LR power ratio obtained by dividing the difference component value of the LR channel by the sum component value, and compares it with a preset threshold thPw. Next, the feature extraction module 72 determines whether or not the LR power ratio is larger than the threshold thPw (step S104).

  When the LR power ratio is larger than the threshold thPw (step S104, YES), the first feature extraction module 72a extracts stereo discrimination information from the stereo signal whose LR power ratio is larger than the threshold thPw (step S105). In the present embodiment, the stereo signal is not a dual monaural signal among signals having two channels, but means a signal having a strong stereo characteristic in which the power ratio of the LR channel is not less than a certain level.

  The first feature extraction module 72a has an LR power ratio (signal amplitude squared sum) in subframe units, a zero-crossing frequency that is the number of times the time waveform of the input audio signal crosses zero in the amplitude direction in subframe units, Discrimination information such as spectral component fluctuations in the frequency domain of the input audio signal is calculated in units of frames. The contents of the discrimination information can be added without being limited to these.

  The first feature extraction module 72a sets a variable paramSet = stereo indicating the discrimination information for stereo with respect to the input audio signal (step S106). The feature extraction module 72 integrates the subframes and extracts frames every several hundred milliseconds (step S107). Next, the feature extraction module 72 obtains a statistical feature amount (for example, average, variance, maximum, minimum, etc.) for each frame from the discrimination information for stereo or the discrimination information for monaural, and generates a feature parameter set (step S108). . The feature extraction module 72 ends the feature amount extraction process.

  If the input audio signal is a dual monaural signal and not a normal stereo signal (NO in step S103), the second feature extraction module 72b receives main / sub selection information determined by the user and becomes a detection target. The channel focus is determined (step S109). The second feature extraction module 72b extracts monaural discrimination information for one of the corresponding channels of main / sub (step S110). Similarly, when the number of channels is not 2 (that is, when the number of channels is 1) (step S102, NO), the second feature extraction module 72b extracts the monaural discrimination information (step S110). Similarly, when the LR power ratio is equal to or less than the threshold thPw (step S104, NO), the second feature extraction module 72b extracts the monaural discrimination information (step S110).

  The second feature extraction module 72a calculates discrimination information such as an LR power ratio, a zero-crossing frequency, and a spectral component variation in units of subframes. The contents of the discrimination information can be added without being limited to these.

  The second feature extraction module 72b sets a variable paramSet = mono indicating the monaural discrimination information for the input audio signal (step S111). Next, the second feature extraction module 72b continues the operation after step S107.

  The discrimination information for stereo and the discrimination information for monaural may be common to each other, or may be unique to each. As the characteristic feature parameter peculiar to the discrimination information for stereo, for example, there is an LR power ratio. The LR power ratio tends to increase in the music section and decreases in the voice section.

  As described above, the feature extraction module 72 extracts the discrimination information for stereo or the discrimination information for monaural according to the contents of the input audio signal together with the channel information of the input audio signal, and the feature amount based on the extracted discrimination information Generate a parameter set. Therefore, the feature extraction module 72 can select optimum discrimination information used for discriminating whether the input audio signal is an audio signal or a music signal. Various feature amount parameter sets generated by the feature extraction module 72 are supplied to the signal type determination module 74.

  Next, the operation of the signal type determination module 74 will be described. FIG. 4 is a flowchart for explaining signal type determination processing using a feature parameter set and channel information. First, the signal type determination module 74 determines whether paramSet = stereo has been set for the input audio signal (step S201). When paramSet = stereo is set (step S201, YES), the first signal type determination module 74a calculates a stereo linear discriminant as follows (step S202).

  The stereo linear discriminant is used to calculate the speech / music identification score S1 used by the signal type determination module 74 to determine whether the input audio signal is a speech signal or a music signal. The signal type determination module 74 assigns a weighting coefficient according to the importance of each feature parameter to the feature parameter set generated by the feature extraction module 72, and takes a linear sum of values multiplied by the coefficient. A voice / music identification score S1 representing the probability of belonging to the voice is calculated. The signal type determination module 74 determines the weighting coefficient by learning using data in which the expected value of the sound type of music / speech is known in advance.

  As this weighting coefficient, a larger value is given to a feature amount parameter that is more effective in determining the signal type. As an example, the signal type determination module 74 uses the following linear determination equation for stereo. In addition, the weighting coefficient is calculated by inputting many known speech signals and music signals prepared in advance as reference data for the speech / music identification score S1, and learning a feature parameter for the reference data. Is done.

The feature parameter set of the kth frame of the reference data to be learned is represented by a vector x, and the signal section {speech, music} to which the input audio signal belongs is represented by y as follows.

Here, each element of the above equation (1) corresponds to the extracted n feature amount parameters. Further, −1 and +1 in the above equation (2) correspond to the speech section and the music section, respectively, and the section that becomes the correct signal type of the speech / music learning data to be used is manually labeled in advance in binary. It is what you did. Since -1, +1 in the above formula (2) is a convenient definition, it may be reversed. Furthermore, the following linear discriminant function is established from the above equation (2).

For k = 1 to N (N is the number of input frames of reference data), a vector x is extracted, and the evaluation value of equation (3) and the error signal sum of equation (2) and equation (4) are minimum. By solving the normal equation, the weighting coefficient β _i (i = 0 to n) for each feature parameter is determined.

  When paramSet = stereo is not set (that is, when paramSet = mono is set) (step S201, NO), the second signal type determination module 74b is similar to the above formulas (1) to (4). Is used to calculate a linear discriminant for monaural (step S202). At this time, unlike the linear discriminant for stereo, the second signal type determination module 74a calculates the linear discriminant for monaural using m feature amount parameters.

  For the linear discriminant for stereo or the linear discriminant for monaural use, the signal type determination module 74 uses the weighting coefficient determined by learning to calculate the evaluation value of the input audio signal that is actually identified for each frame from Equation (3). (Step S204). f (x) corresponds to the voice / music identification score S1.

  Note that the calculation of the speech / music identification score S1 is not limited to the method of multiplying the feature parameter by the weighting coefficient obtained by offline learning using the above-described linear identification function. For example, an empirical threshold is set for the calculated value of each feature parameter, and a weighted score is assigned to each feature parameter in accordance with comparison with the threshold, and a score is calculated. It is possible.

  The signal type determination module 74 determines whether or not S1 <0 (step S205). The signal type determination module 74 determines a music section if S1 <0, and a voice section if f (x)> 0. The signal type determination module 74 determines each frame exclusively as a voice section or a music section.

  When S1 <0 is not satisfied (that is, when it is a voice section) (step S205, NO), the signal type determination module 74 increments the variable cntSp (step S206). When S1 <0 (that is, when it is a music section) (step S205, YES), the signal type determination module 74 increments the variable cntMs.

  The voice / music identification score S1 calculated by the signal type determination module 74 and the incremented variable are supplied to the level calculation module 76. The signal type determination module 74 ends the signal type determination.

  Here, the signal type determination module 74 selects different feature parameter sets depending on whether the input audio signal determined based on the channel information is a stereo signal or a monaural signal. The effectiveness with which the signal type determination module 74 selects the feature parameter set will be described.

  For example, the number n of feature quantity parameters in the stereo feature quantity parameter set is different from the number m of feature quantity parameters in the monaural feature quantity parameter set. As described above, when the input audio signal is a stereo signal, the signal type determination module 74 uses the feature parameter set including the statistical feature calculated from the LR power ratio, which is the discrimination information. An improvement in the detection accuracy of S1 can be expected. On the other hand, when the input audio signal is a monaural signal, the detection accuracy of the speech / music identification score S1 can be improved even if the signal type determination module 74 uses a feature parameter set including the statistical feature calculated from the LR power ratio. I can't expect it. On the contrary, the detection accuracy may be reduced.

Expression (5) is an example in which the first signal type determination module 74a determines the weighting coefficient β _i corresponding to the importance of each feature parameter and is applied to Expression (3). χ n is a feature parameter in the LR power ratio.

  As shown in equation (2), if the value of the linear discriminant function is negative, the musicality of the input audio signal is increased. Here, in the normal stereo music signal, since different musical sounds are arranged in the LR channel, the LR power ratio tends to increase.

  This trend is generally true for any stereo song. As a result of learning, the value of the weighting coefficient corresponding to the feature quantity parameter in the LR power ratio is likely to be relatively larger than the weighting coefficient value in which the other feature quantity parameters indicate the discrimination of the music section / speech section. In other words, the feature amount parameter in the LR power ratio is stronger than the degree to which other feature amount parameters contribute to the determination of the music section / speech section. Therefore, the value of the linear discriminant function also tends to be a large negative value.

  On the other hand, if the input audio signal is music but is a monaural signal, the feature parameter χ n is omitted. The second signal type determination module 74b calculates the value of the linear discriminant function by putting a value of 0 in normal χn. That is, in the value of the linear discriminant function, the term of the feature parameter in the LR power ratio does not contribute to the determination of the music section / speech section. In the second signal type determination module 74b, the detection accuracy of the music section / voice section is lowered. The second signal type determination module 74b determines the weight of the weighting coefficient for each feature parameter in consideration of contribution to the determination of the music section / speech section. The feature quantity parameter in the LR power ratio has a relatively large contribution to the determination of the music section / speech section compared to other feature quantity parameters. When the term of the feature parameter in the LR power ratio is omitted from the linear discriminant function, the second signal type determination module 74b has difficulty in determining the music section / speech section.

  Therefore, the second signal type determination module 74b is a feature parameter set excluding the feature parameter parameter of the LR power ratio (specific to the feature parameter and monaural signal that can be expected to have an effect common to monaural signals and stereo signals). The weighting coefficient value is obtained by the equations (1) to (4).

  The second signal type determination module 74b is more musical than the weighting coefficient value shown in Equation (5) for a specific feature parameter among other feature parameters because there is no feature parameter for the LR power ratio. A coefficient value that strongly indicates is given. Therefore, the second signal type determination module 74b can suppress a decrease in detection accuracy of the music section / speech section.

  As described above, the signal type determination module 74 can prepare an optimum weighting coefficient according to a stereo signal or a monaural signal, and can switch and use a linear determination formula according to channel information of the input audio signal.

  Next, the operation of the level calculation module 76 will be described. FIG. 5 is a flowchart for explaining the level calculation process. The level calculation module 76 can determine that the value is a speech segment if the value of the linear discriminant function obtained by the equation (5) is positive, and a music segment if the value is negative. However, in order to finely control the sound quality of the sound output from the speaker 15 by the control module 63, the level calculation module 76 preferably calculates the value of the linear discriminant function in the form of accuracy information expressed stepwise. Also, with monaural signals, music characteristics do not appear as prominently as stereo signals as feature parameters. Therefore, the musicality score of the linear discriminant function value S1 tends to take a relatively small value. For this reason, the level calculation module 76 may be unstable depending on the music. Therefore, the level calculation module 76 calculates a voice / music level that also serves as score stabilization as follows, for example.

  The level calculation module 76 calculates accuracy information for each of the music section and the voice section based on the linear discriminant function value S1 obtained by the linear discriminant. Here, Sm1 is a music score variable, and Ss1 is a voice score variable. The level calculation module 76 sets Sm1 = −S1 and Ss1 = S1 (step S301). The reason for inverting the sign of S1 in Sm1 is that it is easy to express both voice and music at a positive level.

  The level calculation module 76 calculates the speech / music identification score S1 for each frame for Sm1 (> 0), and continuously counts the number of frames cntMs for which music has been determined in the past. The level calculation module 76 determines whether cntMs is equal to or greater than the specified number of times thNms (step S302).

  When cntMs reaches thNms (step S301, YES), the level calculation module 76 adds a correction score Sm2 (> 0) to be added to Sm1 by step_m (> 0). The level calculation module 76 reduces the correction score Ss2 (> 0) to be subtracted from Ss1 by step_s (> 0). The level calculation module 76 clips the values of Sm2 and Ss2 within a range of appropriate values (min = 0, max = 1, etc.) (step S303).

  As a result, even if the music score variable indicated by Sm1 is a relatively small value, the corrected value of the music score variable is stabilized as time passes.

The level calculation module 76 adds the correction score Sm2 to the music score variable Sm1 as shown in equation (6) (step S304).

The level calculation module 76 subtracts the correction score Ss2 from the speech score variable Ss1 as shown in equation (7) (step S305).

  When cntMs has not reached thNms (step S302, NO), the level calculation module 76 continuously counts the number of frames cntSp for which speech determination has been made in the past for Ss1 (> 0). The level calculation module 76 determines whether cntSp is equal to or greater than the specified number of times thNsp (step S306).

  When cntSp reaches thNsp (step S306, YES), the level calculation module 76 decreases the correction score Sm2 (> 0) to be added to Sm1 by step_m (> 0). The level calculation module 76 adds a correction score Ss2 (> 0) to be subtracted from Ss1 by step_s (> 0). The level calculation module 76 clips the values of Sm2 and Ss2 within a range of appropriate values (min = 0, max = 1, etc.) (step S307).

  Since the level calculation module 76 reduces the correction score Sm2 step by step, the level calculation module 76 has an effect of mitigating sudden correction sound quality fluctuation when changing from music to a voice section.

The level calculation module 76 subtracts the correction score Sm2 from the music score variable Sm1 as shown in equation (8) (step S308).

The level calculation module 76 adds the correction score Ss2 to the voice score variable Ss1 as shown in equation (9) (step S309). The level calculation module 76 can stabilize the voice / music level by adding the correction score Ss2 with the continuity of the determination.

  Next, the level calculation module 76 performs clipping in the range of 0 to 1 in order to appropriately convert Ss1 ′ and Sm1 ′ into a form that can be easily handled in the subsequent stage (step S310). The level calculation module 76 converts Ss1 ′ and Sm1 ′ to a desired resolution level (step S311). The level calculation module 76 converts the music level Lms and the audio level Lsp as N-stage integer values such as 0 to 255, for example.

  The level calculation module 76 performs smoothing in the process of level value conversion (step S312). This is because the level calculation module 76 suppresses a sudden change in voice / music level between frames. That is, when performing smoothing with the past num_fr number of frames, the level calculation module 76 multiplies the sound and music levels of the num_fr number of frames by the weighting factor to obtain a moving average to obtain the final output level (music level Lms , Audio level Lsp). At this time, for example, the level calculation module 76 increases the value of the weighting coefficient by which the voice / music level is multiplied for the most recent past frame.

  The level calculation module 76 can obtain a stable voice / music level with low delay and low load by the above-described score correction and smoothing. The signal type determination module 74 exclusively calculates the music / speech result based on the binary determination result in equation (3). However, since the level calculation module 76 performs score correction / smoothing independently on the voice / music level information, the voice / music level can be calculated as independent values that are not mutually exclusive with time. . The level calculation module 76 outputs the music / sound level as an accuracy corresponding to each sound component in a section such as BGM, for example.

  Further, the level calculation module 76 may control the audio / music level according to the content of the input audio signal to which the detection is applied and the content type to which the input audio signal belongs. For example, if the input audio signal is a monaural signal in which the effect of correcting the music is relatively difficult to obtain compared to the stereo signal, the level calculation module 76 lowers the maximum value of the voice / music level compared to the stereo signal. Set.

  Or, in a drama or variety other than a music program where talk scenes and music scenes appear relatively clearly, various sound effects are likely to be included in the production, and significant fluctuations between the music section and the voice section occur frequently within a short period of time. To do. The level calculation module 76 refers to genre information such as EPG in order to avoid the influence of a sudden change in sound quality due to such fluctuations, and sets the audio / music level of music / audio level to be output low for specific content.

  The sound quality correction module 80 can flexibly control the sound quality correction depending on whether the input audio signal is a music signal or a sound signal, and a stereo signal or a monaural signal. That is, the sound quality correction module 80 performs sound quality correction processing in accordance with the content of the signal using the calculated music / sound level information.

  For example, if the input audio signal is a stereo signal and the music level is high, the sound quality correction module 80 performs correction on the input audio signal that emphasizes a sense of spread such as a surround effect. The sound quality correction module 80 applies equalization center correction to the input audio signal if the input audio signal is a monaural signal and the music level is high. If the input audio signal is a monaural signal and the sound level is high, the sound quality correction module 80 applies contour enhancement with enhanced center localization to the input audio signal. The sound quality correction module 63 performs softer sound enhancement on the input audio signal if the input audio signal is a stereo signal and the sound level is high. Therefore, the sound quality correction module 80 can be easily controlled according to the number of channels of the input audio signal, the height of the voice / music level, and the stability.

  According to the present embodiment, the signal characteristic analysis module 70 can flexibly switch the sound quality correction according to the characteristics of the input audio signal. The signal characteristic analysis module 70 can accurately detect not only stereo signals but also monaural signals. Further, the signal characteristic analysis module 70 can optimally detect an input audio signal having a monaural property or a dual monaural input audio signal even in a stereo signal format. The signal characteristic analysis module 70 can express the accuracy of music / speech by level information after stabilizing instantaneous and local determination blur. Further, the signal characteristic analysis module 70 can calculate the voice / music level with a low delay and a low load based on one discriminant, and obtains it as information that is stabilized and independent between voice and music according to the duration. be able to. As a result, the signal characteristic analysis module 70 can flexibly switch the sound quality correction of the input audio signal according to the classification of monaural / stereo and voice / music.

  The above-described module may be realized by hardware, or may be realized by software using the CPU 64 or the like.

  Note that the present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the scope of the invention in the implementation stage. In addition, the embodiments may be appropriately combined as much as possible, and in that case, the combined effect can be obtained. Further, the above embodiments include inventions at various stages, and various inventions can be extracted by appropriately combining a plurality of disclosed constituent elements. For example, even if some constituent requirements are deleted from all the constituent requirements shown in the embodiment, the problem described in the column of the problem to be solved by the invention can be solved, and the effect described in the column of the effect of the invention Can be obtained as an invention.

  DESCRIPTION OF SYMBOLS 11 ... Digital television broadcast receiver, 15 ... Speaker, 72 ... Feature quantity extraction module, 74 ... Signal type determination module, 76 ... Level calculation module, 80 ... Sound quality correction module

Claims (9)

Discrimination means for discriminating whether the input audio signal is a monaural signal or a stereo signal based on channel information;
A feature parameter set including a plurality of feature amount parameters for discriminating the input audio signal into either an audio signal or a music signal, and is different for the monaural signal or the stereo signal discriminated by the discriminating means Feature extraction means for extracting a feature parameter set;
Audio music indicating whether the input audio signal is close to an audio signal or a music signal using different discriminants according to the feature parameter set for the monaural signal or the stereo signal extracted by the feature extraction means A signal type determining means for calculating an identification score;
Level calculation means for calculating the audio level of the input audio signal and the output level of the music level using the audio music identification score;
Sound quality correction means for performing sound quality correction processing on the input audio signal based on the output level calculated by the level calculation means;
An audio signal correction apparatus comprising:   When the input audio signal is a dual monaural signal, the determination means determines the selected channel of the main or sub channel as a detection target based on selection of either the main channel or the sub channel. Thus, the input audio signal is determined as the monaural signal for the selected channel, or the input audio signal is in the stereo signal format and the LR power ratio of the input audio signal is smaller than a predetermined value. The audio signal correcting apparatus according to claim 1, wherein the input audio signal is discriminated from the monaural signal.   The audio signal correction apparatus according to claim 1, wherein the feature extraction unit extracts an LR power ratio as one of the plurality of feature amount parameters when the input audio signal is the stereo signal.   A plurality of weighting coefficients calculated by learning the plurality of feature parameters using the voice signal and the music signal prepared in advance as reference data for each of the plurality of feature parameters; The audio signal correction apparatus according to claim 1, wherein a sum total obtained by multiplying each of the plurality of feature parameters and the plurality of weighting coefficients is calculated as the speech music identification score.   2. The audio signal correction apparatus according to claim 1, wherein the feature extraction unit divides the input audio signal into a plurality of frames for each predetermined unit and extracts the plurality of feature amount parameters for each of the divided frames.   When the level calculation means determines that the audio music identification score for each of the divided frames calculated by the signal type determination means is the music signal continuously for a predetermined number of times, the level calculation means uses the audio music identification score for music. When the correction score is added so as to increase the correction strength, and the audio music identification score for each of the divided frames calculated by the signal type determination unit is determined to be the audio signal continuously for a predetermined number of times, the audio The audio signal correction apparatus according to claim 5, wherein the correction score is added to the music identification score so as to increase the correction strength for voice.   7. The audio signal correction apparatus according to claim 6, wherein the level calculation means calculates the smoothed output level by taking a moving average of the corrected speech and music identification scores for a plurality of the divided frames.   The level calculation means reduces the maximum value of the output level when the audio signal is the monaural signal, and changes the maximum value of the output level according to the genre of the audio signal. The audio signal correcting apparatus according to claim 7. Based on the channel information, the input audio signal is identified as either a monaural signal or a stereo signal,
A feature parameter set including a plurality of feature amount parameters for discriminating the input audio signal into either an audio signal or a music signal, wherein different feature parameter sets are used for the determined monaural signal or the stereo signal. Extract and
Using a different discriminant according to the feature parameter set for the monaural signal or the stereo signal, calculate a speech music identification score indicating whether the input audio signal is close to a speech signal or a music signal;
Calculating the audio level of the input audio signal and the output level of the music level using the audio music identification score;
Based on the output level, a sound quality correction process is performed on the input audio signal.
Audio signal correction method.

Images