JPH02230895A - Acoustic signal input device - Google Patents

Abstract

PURPOSE:To stabilize sensitivity frequency characteristic by using a level difference caused when data analyzed by a frequency analysis section is compared with a level value in the standard pattern registered in advance in a characteristic control section for each frequency so as to control the gain frequency characteristic of an equalizer with the characteristic control section. CONSTITUTION:An acoustic signal output of a microphone 1 is supplied simultaneously to an equalizer 2 and a frequency analysis section 3. Then the gain frequency characteristic of the equalizer 2 is controlled by the characteristic control section 4 by using a level difference caused when the data analyzed by a frequency analysis section 3 is compared with a level in the standard pattern registered in advance in a characteristic control section 4 for each frequency. Thus, a change in the sound collection distance from a sound source S to the microphone 1 and a change in the transmission frequency characteristic due to the effect of a reflecting body (acoustic transmission distortion) are eliminated to obtain a flat transmission frequency characteristic.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an audio signal input device for a speech recognition device and various audio signal recording devices.

[Conventional technology]

Speech recognition devices and various acoustic signal recording devices use omnidirectional microphones or directional microphones with a first-order gradient or higher sound pressure gradient, depending on the presence or absence of noise in the sound collection environment.

These various microphones have been made smaller, lighter, and higher in performance due to recent advances in electroacoustic conversion material manufacturing technology and LSI technology. However, in the case of directional microphones, as miniaturization progresses, the distance between the acoustic terminals between the front and rear surfaces of the microphone diaphragm decreases.
There is a problem in that when the distance L between the sound source and the microphone changes during actual use, the noise frequency characteristics, especially at low frequencies, change significantly. This is well known as the proximity effect, and an example of its actual measurement is shown in Figure 3. This example is a case of a commercially available tie pin type microphone, and its external dimensions were 9 mmφ x 35 mm. As is clear from Fig. 3, the sensitivity frequency characteristic of this microphone becomes flat when the distance between the sound source and the microphone is approximately 3Qcm (curve S1), and when it is used closer than this distance, the sensitivity at low frequencies decreases. There is a fact that it is rising゛(
Song IS2 is distance IQcm, S3 is 5cm, S4 is 2.5
cm, 35 is 1.25cm SS6 is Qcm). As a result, it is generally experienced that the collected audio signal has a sound quality that is significantly different from the original signal. Also,
It is also true that there are fields in which such changes in frequency characteristics are used adversely to enhance sound quality at low frequencies. However, the fact that the sensitivity frequency characteristics of the microphone during sound collection change due to changes in distance during use is not particularly desirable for voice recognition processing devices, but is also not desirable for general music sound recording. Furthermore, it is also true that the presence of a reflector such as a conference table or a wall in the vicinity of the sound collection microphone or the speaker greatly affects the acoustic transmission characteristics, and in this case it has little to do with the directional characteristics of the microphone. In Figure 4, the distance between the reflector, the speaker, and the microphone is 30 cm, and the distance L between the speaker and the microphone is 10.50.90 cm (curve S7
, S8, S9) shows the transmission frequency characteristics between the speaker microphones. As can be seen, the transmission frequency characteristics change greatly depending on the mutual arrangement of the speaker, the microphone, and the reflector.

[Problem to be solved by the invention]

As a result of recognition processing of voices collected under conditions that caused changes in transmission frequency characteristics as shown in Figures 3 and 4, approximately 20
% reduction in recognition rate. Note that this result is obtained when similar words with a high degree of recognition difficulty are used, so the actual value is likely to be smaller than this, but it is still not within an acceptable range for actual use. .

The present invention has been made in view of the above points, and its purpose is to compensate for changes in the transmission frequency characteristics due to changes in the recording distance between the speaker and the microphone and the effects of reflectors. An object of the present invention is to provide an audio input device capable of obtaining an output signal characteristic such that the output signal level frequency characteristic of the sound is equal to the transmission characteristic when sound is collected in a flat system.

[Means to solve the problem]

In order to solve these problems, the present invention simultaneously supplies the acoustic signal output of the microphone to the equalizer and the frequency analysis section, and also registers the data analyzed by the frequency analysis section in the characteristic control section in advance as a standard pattern. The gain frequency characteristic of the equalizer is controlled by the characteristic control section using the level difference that occurs when the level values for each frequency band are compared for each frequency.

[Effect]

In the acoustic signal human power device according to the present invention, it is possible to obtain flat transmission frequency characteristics by removing changes in transmission frequency characteristics (acoustic transmission distortion) due to changes in sound collection distance from the sound source to the microphone and the influence of reflectors. .

〔Example〕

First, the features of the present invention and the differences from the prior art will be described. The present invention selects and compares the frequency spectrum characteristics of the audio output signal of the microphone with standard patterns of the frequency spectrum characteristics of specific speakers or men and women that have already been stored, and detects changes in the transmission frequency characteristics from the difference. The main feature is that it compensates with an equalizer and attempts to achieve flat sound pickup characteristics. In the prior art, no acoustic input device having such characteristics has been found.

FIG. 1 is a system diagram showing one embodiment of the present invention.

In the figure, 1 is a directional or omnidirectional microphone, 2 is an equalizer for correcting transmission frequency characteristics, 3 is a frequency analysis section, 4 is a characteristic control section, 5 is a signal output terminal, and S is the distance from microphone 1 This is the sound source located at the L position.

Next, the operation of the apparatus shown in FIG. 1 will be explained.

The audio signal a picked up by the microphone 1 is input to an equalizer 2 and a frequency analysis section 3. The audio signal a input to the frequency analysis unit 3 is subdivided into a frequency bandwidth of, for example, 1/3 octave or more, and the audio signal level existing in each band is determined at long time intervals (several seconds). Continuously detected. An example of the analysis result at this time is shown in FIG. 2 by a dotted line S10. In FIG. 2, the horizontal axis is frequency and the vertical axis is long-term effective value level. The value b indicated by the dotted line SIO is transferred to the characteristic control section 4. The characteristic control unit 4 selects a value that has already been registered as a standard audio pattern, for example, the value shown by ACT&iS20 in FIG. 2, and compares it with the characteristic of the dotted line SIO. At this time, the level difference between the solid line and the dotted line is the frequency characteristic deviation that occurs in the sound transmission system between the speaker and the microphone during voice collection. Among the factors that caused this frequency characteristic deviation, in addition to changes in the microphone distance during sound collection and characteristic changes due to reflectors as shown in Figures 3 and 4, there are also characteristics due to differences in the microphones used. It will be evaluated on an overall basis, including changes.

Next, based on the frequency characteristic deviation determined by the characteristic control section 4, data C for correcting the transmission frequency characteristic is sent to the equalizer 2. As a result, a signal output having the same frequency characteristics as if the voice of each speaker was collected using a flat system is outputted to the output terminal 5. Furthermore, since these operations are performed continuously on the time axis, correction of the transmission frequency characteristics can be achieved by following changes in the transmission frequency characteristics that change as the speaker's position changes. becomes.

Note that the following methods can be cited as methods for registering standard speech patterns. The first method is to register a standard pattern for each individual, in which short sentences for learning are uttered in a quiet room with no reflectors, and the sound is collected and analyzed using a microphone with a flat frequency response. This is what we do. Using this standard pattern is effective when the speaker is identified. The second method is to register the average value of the voice spectrum characteristics of the speaker using the speech recognition device.In this case, the characteristics of men and women are particularly different at low frequencies, so they are registered separately. is reasonable. Furthermore, the third method creates a pattern using the average value of audio spectrum characteristics with a high degree of similarity.By preparing multiple characteristic standard patterns, more precise transmission frequency characteristics can be obtained. This achieves flattening of the surface area.

〔Effect of the invention〕

As explained above, the present invention uses the level difference that occurs when the data analyzed by the frequency analysis section and the level value in the standard pattern registered in advance in the characteristic control section are compared for each frequency. By controlling the gain frequency characteristics of the equalizer using the characteristic control section, it is possible to obtain a signal output having the same frequency characteristics as those collected by a flat system for each speaker's voice. This has the advantage of being able to compensate for changes in the sensitivity frequency characteristics that occur in the acoustic pickup system due to changes in the distance between the two or the effects of reflectors. Therefore, since the voice is always collected with a flat human output transmission system ensured, it can be used in audio signal input devices such as voice recognition devices and various recording devices to reduce the factors that reduce the recognition rate. This has the advantage that it is possible to collect sound with clear sound quality without audio distortion.

[Brief explanation of the drawing]

Fig. 1 is a system diagram showing an embodiment of the audio signal input device according to the present invention, and Fig. 2 shows the analysis results of the frequency spectrum characteristics of standard audio (solid line) and the frequency spectrum characteristics of input audio during use (dotted line). Frequency characteristic diagram. Figure 3 is a frequency characteristic diagram showing the change in sensitivity frequency characteristics depending on the distance of a commercially available tie-pin type directional microphone based on actual measurements. Figure 4 is a frequency characteristic diagram showing the change in sensitivity frequency characteristics depending on the distance of a commercially available tie-pin type directional microphone. FIG. 3 is a frequency characteristic diagram showing transmission frequency characteristics. 1...Microphone, 2...Equalizer, 3...
Frequency analysis section, 4...Characteristics control section, 5...Signal output terminal, S...Sound source.

Claims (1)

[Claims] A microphone for picking up audio signals, an equalizer for varying gain frequency characteristics, and dividing the output signal of the microphone into a plurality of frequency bands within a required transmission frequency band and analyzing the signal level for each frequency band. An acoustic signal input device comprising a frequency analysis section and a characteristic control section that has a function of accumulating level distribution data for each frequency band obtained by the frequency analysis section and controls the equalizer, simultaneously supplying an acoustic signal output to the equalizer and the frequency analysis section;
The equalizer is adjusted by using the level difference that occurs when comparing the data analyzed by the frequency analysis section and the level value for each frequency band that has been registered in advance as a standard pattern in the characteristic control section for each frequency. An acoustic signal input device characterized in that a gain frequency characteristic is controlled by the characteristic control section.