JPH05204395A - Audio gain controller and audio recording and reproducing device - Google Patents

Abstract

PURPOSE:To precisely perform a gain control without being affected by noise and intonation contained in audio signals. CONSTITUTION:A nonlinear transforming circuit 41 expands the smaller amplitude signals of audio signals sampled by an audio input device 10, nonlinear transform processes which extend amplitudes and time axes and the generation probability of amplitudes is averaged. An amplitude deciding circuit 42 computes an optimum amplitude for the gain control by these signals, a gain deciding circuit 43 decides the gain corresponding to the calculated amplitude and a multiplier 30, to which the audio signals from the device 10 through an audio CODEC 20, is given as amplitude coefficients. Then, optimum coefficients are multiplied to the audio signals and an optimum gain control is performed. The audio gain controller 20 has a nonlinear transforming section 21d which nonlinear transforms amplitudes so as to average the generation frequency of inputted amplitudes and obtains an optimum gain as to reduce the adverse effects caused by noise and conversation intonation. The gain is multiplied to the amplitudes of audio signals decoded by the voice codic 20 at the mutliplier 30.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention relates to a voice codec (CO
DEC: Synthetic word of coder and decoder) for automatic gain control for voice, particularly for voice that automatically controls the gain of a voice signal input from a voice input device such as a microphone to an appropriate value. The present invention relates to an automatic gain control device and an audio recording / reproducing device equipped with the audio automatic gain control device.

[0002]

2. Description of the Related Art Generally, a voice codec is a device having both a function of encoding voice which is an analog signal and a function of decoding encoded data. An audio recording having a configuration in which an analog audio signal input through an audio input device is converted into encoded data by such an audio codec, and then the encoded data is recorded on a recording medium such as a tape cartridge. There is a playback device.

In some telephones with an answering machine function, a voice recording / reproducing apparatus using a semiconductor memory device is adopted instead of the voice recording / reproducing apparatus using a magnetic tape.
In an audio recording / reproducing apparatus using such a semiconductor memory device, an audio signal is encoded and temporarily stored in this semiconductor memory device, and then this is decoded and reproduced as audio.

In order to improve the reproduced sound quality, it is necessary to control the gain of the audio signal to an appropriate value by the audio gain control device so that the amplitude of the audio signal becomes an appropriate size. Conventionally, this kind of gain control has been performed by feedback control in which an averaged voice amplitude or voice power obtained by monitoring (statistical processing) a voice signal is fed back as a gain coefficient. For example, as a method for extracting the characteristics of a voice signal, there is a method based on frequency distribution by a statistical method.

[0005]

However, when a voice signal is observed for a long time, the frequency distribution has a distribution in which the probability of occurrence of a voice signal of small amplitude is high and the probability of occurrence of a voice signal of large amplitude is low. , Not uniform distribution. Therefore, according to the frequency distribution method, the nature of the voice cannot be accurately grasped, and the accurate amplitude coefficient cannot be obtained, so that the sound quality is deteriorated. In particular, in a short conversation, the voice cannot be limited to a specific property, and it is difficult to obtain appropriate amplitude information, which becomes a larger problem. In addition, in the case of conversation, there is a problem that it is easily affected by intonation.

If the voice gain control device performs the gain control while feeding back the voice signal in accordance with the amplitude of the voice signal, the voice becomes monotonous with little intonation. However, in a voice recording / reproducing apparatus used for a telephone with an answering machine, etc., it is possible to decide how to adjust the gain afterwards because the recording time and the reproducing time of the voice signal are different. It is possible. Therefore, an audio gain control device has been already developed, which divides an audio signal into predetermined intervals, calculates the gain adjustment amount based on the average value of the amplitudes in this interval, and records this together with the digital data of the audio signal. There is. By performing the gain control for each predetermined section of the audio signal in this manner, it is possible to save the intonation of the audio during that period. However, the sound of a normal spoken word has a voiced part in which a word is uttered and a silent part in each word segment. Then, in the spoken language having many silent portions, the proportion of the silent portions in the predetermined section increases, and the average value of the amplitude of the audio signal becomes abnormally low. For this reason, the conventional voice gain control device has a problem in that when the voice signal is divided into predetermined intervals and the gain is adjusted, it is impossible to properly adjust the gain for a voice having many silent portions. It was In addition, if the section that divides the audio signal is shortened in order to solve this problem, the intonation will be poor and the problem will not be solved, as in the case where the amplitude is fed back at any time.

Further, in the above-mentioned conventional audio recording / reproducing apparatus, since the audio signal is gain-controlled and encoded and then stored in the recording medium, only the audio signal after the gain control is always decoded during reproduction. become. For this reason, the conventional voice recording / reproducing apparatus has a problem in that the reproduced voice has a monotone in volume and intonation and cannot accurately convey the nuance of the recorded message.

It is an object of the present invention to provide a voice gain control device capable of performing accurate gain control without being affected by noise and intonation included in a voice signal.

Another object of the present invention is to create a frequency distribution of amplitude for a voice signal in a predetermined section, extract only a voiced part from this frequency distribution, and calculate an average value thereof, thereby eliminating the influence of a silent part. It is an object of the present invention to provide a voice gain control device capable of performing gain control without being affected.

Still another object of the present invention is to perform gain control of a voice signal for each predetermined section, and to record a gain coefficient for this gain control separately from digital data of the voice signal. An object of the present invention is to provide an audio recording / reproducing apparatus capable of reproducing an audio without impairing the volume and intonation of audio while suppressing an increase in recorded data to a minimum.

[0011]

A voice gain control device according to the present invention is a voice gain control device for automatically controlling the gain of a voice signal input from a voice input device, and the amplitude of the voice signal. Of a small signal, a non-linear conversion circuit for averaging the occurrence probabilities of the small-amplitude signal and the large-amplitude signal, and an amplitude value for gain control based on the voice signal averaged by the non-linear conversion circuit. The above object is achieved by including a gain determination circuit that calculates and determines the gain based on the calculated value, and controls the amplitude of the audio signal by the determined gain.

The non-linear conversion circuit and the gain determination circuit are constituted by a digital signal processor provided with a multiplier, a ROM, a RAM and an arithmetic unit, and the arithmetic processing contents of the non-linear conversion circuit and the gain determination circuit are performed by software. May be.

The software operation processing of the non-linear conversion circuit may be performed by the μ-law compression characteristic method which is one of the non-linear quantization methods.

In the software operation processing of the gain determination circuit, the input level given to the gain determination circuit from the non-linear conversion circuit is compared with a preset maximum value of the input level, and the input level is determined as follows. When it is larger than the maximum value, the maximum value may be adopted as an input level, and the gain may be determined according to the maximum value by a software processing procedure.

The input signal from the non-linear conversion circuit may be filtered using the multiplication function of the digital signal processor.

The audio gain control device of the present invention is an audio gain control device used in an audio recording / reproducing device for encoding an audio signal and recording the encoded audio signal in a storage device. Frequency distribution creating means for creating a distribution, voiced part extracting means for extracting only voiced parts from the frequency distribution created by the frequency distribution creating means, and frequency distribution of voiced parts extracted by the voiced part extracting means And an average value calculating means for calculating an average value of the amplitude and a gain determining means for determining a gain based on the average value calculated by the average value calculating means, thereby achieving the above object. ..

An audio recording / reproducing apparatus of the present invention is an audio recording / reproducing apparatus which encodes an audio signal and records it in a storage device, and decodes and reproduces the audio signal. The audio signal is encoded and recorded in the storage device. Encoding means, and an average value calculating means for calculating the average value of the amplitude of each predetermined section of the audio signal,
Gain coefficient calculation means for calculating a gain coefficient from the average value calculated by the average value calculation means and recording the gain coefficient in the storage device; decoding means for reading and decoding the encoded signal from the storage device; Gain control means for reading the gain coefficient calculated by the gain coefficient calculation means from the storage device for the voice signal decoded by the decoding means, and adjusting the gain based on the gain count; and The above object is achieved by the above.

[0018]

1 is a block diagram of a system in which a voice gain control device of the present invention is used in a voice recording / reproducing device equipped with a voice codec.

The voice input device 10 monitors a series of voice signal trains and samples the amplitude of the voice signal trains at a predetermined sampling pitch. The sampled signal is input to the audio codec 20. The audio codec 20 encodes the input signal into encoded data, decodes the encoded data, and supplies the decoded data to the multiplier 30. More specifically, the encoded data encoded by the encoding circuit in the audio codec 20 is output to, for example, the recording unit of the audio recording / reproducing apparatus, and then recorded on a recording medium (not shown) such as a tape cartridge. To be recorded. As a result, the conversation content and the like are recorded. The encoded data read from the recording medium is
It is decoded by the decoding circuit in the audio codec 20 and given to the multiplier 30.

The multiplier 30 includes an audio gain control device 40.
The gain coefficient determined by is input, and a signal obtained by multiplying the decoded signal by the gain coefficient, that is, a signal having an appropriate amplitude for which gain control is performed is output.

The audio gain control device 40 includes a non-linear conversion circuit 41 to which the audio signal sampled by the audio input device 10 is input, and an amplitude calculation circuit to which the signal processed by the non-linear conversion circuit 41 is input. 42, and a gain determination circuit 43 that calculates the gain coefficient from the amplitude value for gain control calculated by the amplitude calculation circuit 42 and gives the calculated gain coefficient to the multiplier 30.

An important feature of the present invention is that the nonlinear conversion circuit 4
The amplitude is calculated by the amplitude calculation circuit 42 based on the audio signal nonlinearly converted by 1. As the circuits themselves of the non-linear conversion circuit 41, the amplitude calculation circuit 42 and the gain determination circuit 43, known circuits existing in the past may be used.

The non-linear conversion circuit 41 of this embodiment performs non-linear conversion of the input audio signal with the following contents. That is, for a signal having a small amplitude in the audio signal, a non-linear conversion process for expanding the amplitude of the signal or the time axis is performed. As a result, the occurrence probabilities of the small amplitude signal and the large amplitude signal are averaged. Therefore, a uniform frequency distribution can be obtained by statistically processing the voice signal to obtain its frequency distribution.
In addition, according to such an averaging process, noise with a large signal amplitude included in the audio signal is compressed, so that the adverse effect of the noise can be eliminated. Therefore, if the gain is controlled based on the averaged audio signal, proper gain control with less influence of noise and intonation becomes possible.

The amplitude calculation circuit 42 is a nonlinear conversion circuit 41.
The amplitude required for gain control is calculated based on the amplitude of the audio signal whose occurrence probability is averaged by. Various methods are known as methods for calculating the amplitude. A method of obtaining the maximum value of the amplitude of the signal input from the non-linear conversion circuit 41 for each predetermined period and setting the maximum value of the amplitude of each period as the “amplitude” for each period, or the frequency within the predetermined period A method is known in which an average amplitude is obtained on the basis of a distribution, and the average amplitude is used as “amplitude” for each period.

The calculated amplitude data is used in the gain determination circuit 4
The gain corresponding to the amplitude value calculated here is determined (calculated) and output to the multiplier 30 as an amplitude coefficient.

In parallel with the amplitude calculation required for the above gain control, the voice codec 20 performs an encoding process. That is, the present embodiment has a system configuration in which the gain is determined in parallel with the encoding process so as not to affect the sound quality of the voice codec.

The voice gain control device 40 of this embodiment, which comprises the non-linear conversion circuit 41, the amplitude calculation circuit 42, and the gain determination circuit 43, can be realized by an analog circuit.
If the voice gain control device 40 is configured with a DSP (Digital Signal Processor) and the software executes the arithmetic processing of the non-linear conversion circuit 41, the amplitude calculation circuit 42, and the gain determination circuit 43, the arithmetic processing is performed quickly and accurately. There is an advantage that can be done. The details will be described below.

FIG. 2 shows the system configuration of the DSP. D
Multiplication and addition are frequently performed in the SP digital signal processing. As hardware for this, the DSP has
A parallel multiplier (hereinafter simply referred to as a multiplier) 104 is provided. The multiplication in the multiplier 104 is called multiplication for each row (bit) of the multiplier with the multiplicand (AND operation) to calculate a partial product, and the partial products are added to obtain a final product. Done in shape. The parallel multiplier 104 accelerates the processing by calculating the partial products and adding the partial products in parallel.

Here, the arithmetic processing executed by the DSP is characterized in that there are many so-called product-sum operations for accumulating multiplication results one after another. Therefore, the DSP arithmetic circuit, that is, A
The LU 100 is designed so that accumulation of multiplication results by the multiplier 104 can be continuously executed in each cycle. That is, the multiplication result of the multiplier 104 is directly calculated by the ALU1.
00, and the operation result of the ALU 100 is an accumulator 10 which is a register for accumulation.
5 (stored) and accumulator 10
The output of 5 is given to the ALU 100 (see FIG. 2). The output of the accumulator 105 is given to the multiplier 104 via the selector 103.

As described above, the DSP can perform multiplication and addition at high speed, but it is necessary to successively provide data to the ALU 100 in order to bring out the processing capacity of the ALU 100 and increase the throughput of the ALU 100. For this reason, the DSP is provided with a plurality of data buses 107 (only one set of data buses is shown in the illustrated example). The data bus 107 shown in the drawing is a ROM 101, a RAM 102 in which data is written / read at a corresponding address according to an address given from the address bus 106.
Functions as a bus that connects the output of the above-mentioned and the input register of the multiplier 104. That is, the data bus 107 is a bus for setting a multiplier and a multiplicand for the multiplier 104 every cycle.
02, data is continuously read from the ROM 101 and multiplication is performed.

Next, referring to FIGS. 3 and 4, the ALU
The contents of the non-linear conversion processing by 100 will be described.
Here, this non-linear conversion processing is similar to the processing content of the above-mentioned non-linear conversion circuit 41, and is a program for the non-linear conversion processing stored in the ROM 101, that is, the ROM.
This is performed by the ALU 100 executing an algorithm according to the program data stored in 101. More specifically, the non-linear conversion process is performed according to the μ-law compression characteristic method, which is one of the non-linear quantization methods. In this non-linear conversion process, y (compressed signal level) in the conversion formula shown in the following formula (1) is calculated, and this y is set as a uniform frequency distribution obtained by averaging the input level from the voice input device 10. obtain.

Y = (x−a) · (16 / b) + c (1) where x: pre-compression signal level (input level from the voice input device 10) a, b, c: of the voice input device 10 It is a coefficient that matches the characteristics.

The procedure for calculating y will be described in more detail below with reference to FIGS. 3 and 4. First, when this program starts, the ALU 100 sets 0, 32, 0 to the coefficients a, b, c for obtaining the conversion formula of the above formula (1) in step S1. Succeedingly, in a step S2, it is determined whether or not the value of the input level x from the voice input device 10 is larger than 31, and if it is confirmed that the value is smaller, the process proceeds to a step S16 shown in FIG. Set to the input level. That is, in step S16, (x-a) in the above formula (1) is calculated.

On the other hand, in step S2, the input level x is 31.
If it is determined that the above is the case, the setting process of the input level x is not performed in this case, and then the coefficients a, b, c
Are set to 31, 64, and 16, respectively. Then, in step S4, it is determined whether or not the input level x is higher than 95, and if it is confirmed that the input level x is lower than 95, the process proceeds to step S16 in the same manner as described above, where the input level x-a at this time is set to the input level x. To do.

If it is determined in step S4 that the input level x is 95 or more, then in step S5, the coefficients a, b, c are calculated.
Are set to 95, 128, and 32, respectively, and step S6
Then, it is determined whether or not the input level x is larger than 223.
If you confirm that the input level x is smaller than 223,
Similarly to the above, the process proceeds to step S16, where the input level x
Is set, and if it is 223 or more, step S
In step 7, new coefficients a, b and c are set. In the same manner, the processes of steps S7 to S15 are performed, and the process proceeds to step S16 to input level x-a corresponding to the coefficient a.
To set.

When the input level x is set in step S16, the ALU 100 next sets (16 / b) to the input level (x-a) calculated in step S16 in the multiplier 104 in step S17. (X-a) in the above equation (1) calculated by multiplying by
Accumulator 105 stores x (16 / b) as a new input level.

Subsequently, in step S18, the input level (x−a) × (16) accumulated in the accumulator 105 is stored.
/ B) is read out and the input level (x−
a) × (16 / b) is added to the coefficient c, and this value (x−a) × (16 / b) + c is finally obtained as the input level, that is, the signal level y after conversion, The accumulated value is stored in the accumulator 105.

Next, according to FIG. 6, the amplitude calculation circuit 42 described above is used.
The contents of the process corresponding to the calculation process of are described. A
When the program for amplitude calculation stored in the ROM 101 starts, the LU 100 determines whether or not the value of the input level x is larger than the default value in step S61. Set to maximum value. On the other hand, if it is determined in step S61 that the input level is less than or equal to the default value, in this case, the maximum value setting process is not performed, and the process proceeds to END. Thus, by repeating the same processing as the above processing for a plurality of input levels sequentially input from the non-linear conversion circuit to the amplitude calculation circuit, the maximum value of the input level (output of the non-linear conversion circuit 41) in a predetermined period. Can be obtained. In this example, the maximum value of the input level is obtained and used as the "amplitude" of each period. However, the multiplication function of the DSP can be used to obtain the amplitude in various ways. For example, not to calculate noise as "amplitude",
It is also possible to filter the input level and remove noise from the input level when calculating the amplitude.

As described above, by expanding the signal having a small amplitude in the audio signal by the non-linear conversion circuit and expanding the amplitude and time axis of the signal, the occurrence probability of the audio signal amplitude can be averaged. Then, by this averaging, noise having a large signal amplitude included in the audio signal is compressed. Therefore, if the gain is controlled based on the averaged audio signal, proper gain control with less influence of noise and intonation can be performed.

FIG. 7 shows the configuration of another voice gain control apparatus according to the present invention. According to this audio gain control device, when an audio signal is input from the audio input device 1, first, the frequency distribution creating means 2 creates an amplitude frequency distribution of the audio signal in each predetermined section. This frequency distribution is
It is a histogram created by subdividing a predetermined section (for example, a sampling period) of an audio signal, quantizing the amplitude of the audio signal for each period, and counting the number of periods in which the quantized values match.

Next, the voiced part extraction means 3 extracts only the voiced part from this frequency distribution. Since the frequency distribution represents the frequency for each amplitude level of the audio signal in the predetermined section, the frequency distribution of only the voiced part can be extracted by discarding the frequency below a certain amplitude level. ..

Then, the average value calculating means 4 calculates the average value of the amplitudes from the frequency distribution obtained by extracting only this voiced portion, and the gain determining means 5 determines the gain based on this average value. Therefore, if the level of the voiced portion of the audio signal is constant, a constant average value can be obtained regardless of the number of silent portions, and thus the gain based only on the voiced portion can be determined. ..

As a result, according to the audio gain control apparatus of the present invention, the audio gain control apparatus is not affected by the silent part of the audio signal,
It becomes possible to always perform optimum gain control.

FIG. 8 shows the configuration of an audio recording / reproducing apparatus equipped with the audio gain control apparatus of the present invention.

The voice signal input from the voice input device 10 is sent to the encoding circuit 12 and the frequency distribution creating circuit 13a. The encoding circuit 12 uses a normal PCM (pulse code m).
Although it may be a circuit for converting an audio signal into digital data by an odulation method, it is preferable to use an ADPCM (adaptive differentia) in order to enable recording for a long time.
l PCM) is a circuit that compresses encoded data using the method. The digital data encoded by the encoding circuit 17 is later decoded by the decoding circuit 16. The decoding circuit 16 is a circuit that decodes the digital data encoded by the encoding circuit 12 into an audio signal by inverse conversion of the same method. And the decoding circuit 1
The audio signal decoded by 6 is an amplitude adjusting circuit 19
Is sent to the audio output device 18 and is reproduced as audio. The audio codec of this embodiment includes an encoding circuit 12 and a decoding circuit 16.

The frequency distribution creating circuit 13a is a circuit which samples and quantizes the input audio signal and creates a frequency distribution of the quantized amplitude for each period of a predetermined sampling number. This frequency distribution is a histogram created by counting the number of sampling periods in which the amplitudes of the quantized audio signals match within a predetermined number of sampling periods of the audio signals. The output of the frequency distribution creating circuit 13a is
It is connected to the voiced part extraction circuit 13b. The voiced part extraction circuit 13b is a circuit for extracting only the voiced part from the frequency distribution created by the frequency distribution creation circuit 13a. Also,
The output of the voiced part extraction circuit 13b is the average value calculation circuit 13
connected to c. The average value calculation circuit 13c is a circuit that calculates the average value of the amplitude levels within the period of the predetermined sampling number based on the input frequency distribution. Then, the gain determination circuit 14 later determines the gain of the audio signal based on the average value of the amplitude levels calculated by the average value calculation circuit 13c.

The output of the gain determining circuit 14 is connected to the amplitude adjusting circuit 19 via the gain control circuit 17. The gain control circuit 18 is a circuit that controls the amplitude adjustment circuit 19 based on the gain determined by the gain determination circuit 14 and adjusts the gain of the audio signal decoded by the decoding circuit 16.

The frequency distribution creating circuit 13a, the voiced part extracting circuit 13b, and the average value calculating circuit 13c of this embodiment are shown in FIG.
It can also be used as the amplitude calculation circuit 42 of the embodiment shown in FIG.

The operation of the audio recording / reproducing apparatus equipped with the audio gain control device having the above-described structure will be described below.

The voice signal from the voice input device 10 is input to the encoding circuit 12 and the frequency distribution creating circuit 13a.
The audio signal sent to the encoding circuit 12 is encoded into digital data by the ADPCM system or the like, and is temporarily stored in a semiconductor memory device (not shown) as appropriate.

The frequency distribution of the voice signal sent to the frequency distribution creating circuit 13a is first created every predetermined sampling period. Next, this frequency distribution is sent to the voiced part extraction circuit 13b to extract only the voiced part. Since the frequency distribution represents the frequency for each amplitude level of the audio signal within the period of the predetermined sampling number, the frequency distribution of only the voiced part is extracted by truncating the frequency below a certain amplitude level. be able to. Then, the frequency distribution obtained by extracting only the voiced portion is sent to the average value calculation circuit 13c, the average value is calculated, and the average value is temporarily stored in a semiconductor memory device (not shown). Therefore, if the level of the voiced portion in the audio signal is constant, a constant average value can be obtained regardless of the number of silent portions.

The digital data stored in the semiconductor memory device as described above is sequentially read and decoded by the decoding circuit 16 into a voice signal. Then, this audio signal is sent to the audio output device 18 via the amplitude adjusting circuit 19 and reproduced as audio. Further, the average value data stored in the semiconductor memory device is read one by one each time the digital data is read by a predetermined sampling number and sent to the gain determination circuit 14. The gain determination circuit 14 determines the gain based on the average value data and sends it to the gain control circuit 17 to adjust the gain of the audio signal passing through the amplitude adjustment circuit 19.

As a result, the voice reproduced from the voice output device 18 can be always subjected to the optimum gain control based on only the voiced portion without being affected by the large amount of the silent portion in the original voice. Like

FIG. 9 shows the structure of a voice recording / reproducing apparatus according to the present invention. The audio signal input from the audio input device 21 is data-compressed and encoded by the encoding means 22 as required. The average value calculating means 23 calculates the average value of the amplitude of each predetermined section, and the gain determining (gain coefficient calculating) means 24 calculates the gain coefficient of this audio signal. The gain coefficient calculated in this way is stored in the semiconductor memory device 25 together with the digital data for each predetermined section encoded by the encoding means 22.

When reproducing this audio signal, the decoding means 26 first converts the digital data into the semiconductor memory device 25.
Read from and decrypt. Further, the gain control means 27
Each time the digital data in the predetermined section is decoded, the gain coefficient is read from the semiconductor memory device 25, and the gain of the decoded audio signal is adjusted based on the gain coefficient.

As a result, according to the audio recording / reproducing apparatus of the present invention, the audio signal gain can be adjusted and output from the audio output device 28, as in the conventional case. Moreover,
In the case of the present invention, since the gain coefficient is constant over a predetermined section, it is possible to store and reproduce the voice intonation within this section to some extent. Furthermore, since this gain coefficient is generated for each predetermined section of voice, even if it is added to the digital data of the voice signal, the data amount does not increase significantly. Further, if the gain coefficient of the gain control means 27 is forcibly fixed, it is possible to reproduce the original voice with the original volume and intonation unchanged.

FIG. 10 is a block diagram showing the structure of a voice recording / reproducing apparatus of the present invention used in a telephone having an answering machine function. The voice signal input from the voice input device (telephone line) 10 is the contact point 11 a and the contact point 1 of the switch 11.
It is switched to and sent from the two circuits by 1b. The circuit on the side of the contact 11a of the switch 11 is the contact 9a of the switch 9.
It is connected to the handset 18b via. The circuit on the side of the contact 11b of the switch 11 is branched into two and connected to the inputs of the encoding circuit 12 and the average value calculating circuit 13d.

The encoding circuit 12 uses a normal PCM (pulse
A circuit for converting an audio signal into digital data by a code modulation) method may be used, but in general, ADPCM (adaptive different) is used to enable long-time recording.
A circuit that compresses the encoded data by the ial PCM) method is used. The digital data encoded by the encoding circuit 12 is sequentially stored in the semiconductor memory device 15. Further, the digital data stored in the semiconductor memory device 15 is sequentially read by the decoding circuit 16.
The decoding circuit 16 is a circuit that decodes the digital data encoded by the encoding circuit 12 into an audio signal by inverse conversion of the same method.

The average value calculation circuit 13d is a circuit for sampling the input audio signal and calculating the average value of the amplitudes thereof for each period of a predetermined sampling number. It should be noted that the sampling of the audio signal can be performed together with the sampling in the encoding circuit 12.
The average value calculated by the average value calculation circuit 13d is sent to the gain determination circuit 14. The gain determination circuit 14 is a circuit that compares the average value calculated by the average value calculation circuit 13d with a predetermined reference amplitude value and calculates a gain coefficient from the ratio. The gain coefficient calculated by the gain determination circuit 14 is also
The data is sequentially stored in other areas of the semiconductor memory device 15. However, since only one data is generated for this gain coefficient for each of a plurality of sampling periods of the audio signal, the encoding circuit 1
It occupies only a small storage capacity as compared with the digital data output from 2. The gain coefficient stored in the semiconductor memory device 15 is read by the gain control circuit 17. The gain control circuit 17 is a circuit that adjusts the gain of the audio signal to be reproduced by multiplying the audio signal output from the decoding circuit 16 by the gain coefficient read out every predetermined period of the audio signal. The output of the gain control circuit 17 is also transmitted through the contact 9b of the switch 9 to the receiver 1
It is connected to 8b.

The operation of the audio recording / reproducing apparatus having the above configuration will be described.

During a normal call, the switch 11 contacts the contact 1.
The switch 9 is connected to the contact 9a side. Therefore, the voice signal from the telephone line 10 is sent to the handset 18b as it is, and a normal call can be performed.

When the answering machine function is working, the switch 11 is connected to the contact 11b side. Therefore, the voice signal from the telephone line 10 is first transmitted to the encoding circuit 1
2 and the average value calculation circuit 13d.

The audio signal sent to the encoding circuit 12 is A
Is encoded into a digital signal by the DPCM method,
The data is sequentially stored in the semiconductor memory device 15. The average value of the amplitude of the audio signal sent to the average value calculation circuit 13d is calculated for each period of a predetermined sampling number. Then, this average value is sent to the gain determination circuit 14 and compared with a predetermined reference amplitude value, a gain coefficient is calculated from the ratio, and this gain coefficient is calculated in the semiconductor memory device 15.
Are sequentially stored in other areas.

When reproducing the telephone message recorded during the absence as described above, the switch 11 is connected to the contact 21b side. Then, the decoding circuit 16 sequentially reads the digital data from the semiconductor memory device 15 and decodes it into an audio signal. At this time, the gain control circuit 17 also sequentially reads the gain coefficient from another area of the semiconductor memory device 15 for each predetermined period of the decoded audio signal, and multiplies the gain coefficient by the audio signal in the corresponding period. Adjust the gain by. Therefore, the voice signal output to the handset 18 via the switch 22 is gain-controlled. However, since this gain coefficient is constant for a plurality of sampling periods, the intonation of speech within this period is preserved to some extent.

If the user wants to hear the above message without controlling the gain, the gain coefficient of the gain control circuit 17 is fixed without being read from the semiconductor memory device 15. Then, the audio signal decoded by the decoding circuit 16 can be reproduced while maintaining the original audio volume and intonation.

The frequency distribution creating circuit 13 shown in FIG. 8 is used instead of the average value calculating circuit 13d of this embodiment.
a, voiced part extraction circuit 13b, and average value calculation circuit 13c
May be used.

[0067]

As described above, according to the voice gain control apparatus of the present invention, the non-linear conversion circuit performs the non-linear conversion processing for averaging the occurrence probabilities of the amplitude of the voice signal, and then the averaged signal is used. Since the amplitude required for gain control is obtained and the gain of the audio signal is controlled by the gain corresponding to the obtained amplitude, the adverse effect of noise and intonation can be eliminated. Therefore, the reliability of the gain control can be improved, and when applied to a voice recording / reproducing apparatus equipped with a voice codec, there is an advantage that the sound quality can be remarkably improved. Further, the optimum gain control can always be performed based on only the voiced portion without being affected by the large proportion of the silent portion occupied in the audio signal.

Further, it becomes possible to reproduce the sound without impairing the loudness and the intonation of the sound, while suppressing the increase of the recorded data to the minimum.

[Brief description of drawings]

FIG. 1 is a block diagram when a voice gain control device of the present invention is used in a voice recording / reproducing device equipped with a voice codec.

FIG. 2 shows a configuration of a Digital Signal Processor.

FIG. 3 is a flowchart showing a part of conversion operation of the nonlinear conversion circuit used in the embodiment of the present invention.

FIG. 4 is a flowchart showing the rest of the conversion operation performed in the embodiment.

FIG. 5 is a graph showing a conversion relationship of non-linear conversion performed in the example.

FIG. 6 is a flowchart showing the operation of the amplitude calculation circuit used in the embodiment of the present invention.

FIG. 7 shows a configuration of a voice gain control device according to the present invention.

FIG. 8 shows a configuration of an audio recording / reproducing apparatus including an audio gain control device of the present invention.

FIG. 9 shows a configuration of a voice recording / reproducing apparatus according to the present invention.

FIG. 10 is a block diagram showing a configuration of a voice recording / reproducing apparatus of the present invention used for a telephone with an answering machine function.

[Explanation of symbols]

 10 voice input device 20 voice codec device 41 non-linear conversion circuit 42 amplitude calculation circuit 43 gain determination circuit

Claims (7)

[Claims] 1. A voice gain control device for automatically controlling the gain of a voice signal input from a voice input device, wherein the amplitude of a voice signal having a small amplitude is extended, and the amplitude of the voice signal having a small amplitude is compared with that of the voice signal. A non-linear conversion circuit that averages the occurrence probabilities of large signals, and a gain determination circuit that calculates an amplitude value for gain control based on the audio signal averaged by the non-linear conversion circuit and determines the gain based on the calculated value. A gain control device for voice, which controls the amplitude of the voice signal according to the determined gain. 2. The non-linear conversion circuit and the gain determination circuit are constituted by a digital signal processor provided with a multiplier, a ROM, a RAM and an arithmetic unit, and the arithmetic processing contents of the non-linear conversion circuit and the gain determination circuit are realized by software. The audio gain control device according to claim 1, wherein 3. The audio gain control device according to claim 2, wherein the software operation processing of the non-linear conversion circuit is performed by a μ-law compression characteristic method which is one of non-linear quantization methods. 4. A software operation process of the gain determination circuit, which compares an input level given to the gain determination circuit from the non-linear conversion circuit with a preset maximum value of the input level to obtain the input level. Is larger than the maximum value, the gain control device for voice according to claim 2, wherein the maximum value is adopted as an input level and the gain is determined according to the maximum value by a software processing procedure. 5. The audio gain control device according to claim 2, wherein the input signal from the non-linear conversion circuit is filtered by using the multiplication function of the digital signal processor. 6. A voice gain control device used in a voice recording / reproducing device for encoding a voice signal and recording it in a storage device, wherein a frequency distribution creating means for creating a frequency distribution of amplitude in a predetermined section of the voice signal. And a voiced part extracting means for extracting only a voiced part from the frequency distribution created by the frequency distribution creating means, and an average value of the amplitudes thereof is calculated from the frequency distribution of the voiced parts extracted by the voiced part extracting means. A gain control device for voice, comprising: an average value calculating means for controlling the gain and a gain determining means for determining a gain based on the average value calculated by the average value calculating means. 7. An audio signal is encoded and recorded in a storage device,
An audio recording / reproducing apparatus for decoding and reproducing the audio signal, comprising: an encoding means for encoding an audio signal and recording it in the storage device; and an average value calculation for calculating an average value of amplitudes of the audio signal for each predetermined section. Means, gain coefficient calculation means for calculating a gain coefficient from the average value calculated by the average value calculation means and recording it in the storage device, and decoding means for reading and decoding the encoded signal from the storage device And gain control means for reading the gain coefficient calculated by the gain coefficient calculation means from the storage device for the voice signal decoded by the decoding means, and adjusting the gain based on the gain count. Audio recording / playback device.