JPH0816199A - Sound recording device - Google Patents

Abstract

PURPOSE:To conduct a high quality sound recording of digital voice signals for a long time. CONSTITUTION:Input voice signals are converted into gidital data in an A/D converter 12, coded by a coder 14 and stored in a voice memory 30. Then, the inputted signals are fed to a characteristic discrimination section 24 and the distinction of the sex of the speaker is made by checking the frequency characteristics of the data, for example, by discriminating the level of the voice signal of lower frequency. The discrimination result is supplied to a control section 26, which sets the sampling frequency of the converter 12 lower than that of a female when the speaker is discriminated to be a male. Since a male voice contains much lower frequency components, a lower sampling frequency does not adversely affect the tone quality. By conducting the above control, an efficient data compression is accomplished and a long time recording is performed while keeping the tone quality high.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a recording device for recording a voice signal as digital data.

[0002]

2. Description of the Related Art Conventionally, a semiconductor memory has been used as a recording medium for digital data, and it has also been used for recording voice signals such as recording on an answering machine. Such recording is performed by, for example, digitizing an audio signal by an A / D converter and then DSP (digital signal processing).
It is performed by high-efficiency encoding by processing and recording in a large-capacity LSI memory. For reproduction, the coded data read from the memory is DSP processed and decoded, and then D /
Return to audio signal with A converter.

In such a system, the sampling frequency for A / D conversion is usually fixed at 8 KHz, but in order to enable recording for a longer time with a limited memory capacity, 6 kHz, 4 kHz. 3 kHz is also selectable. Also, the number of bits (quantization bit number) of encoded data is often fixed to any one of 1 to 4 bits, but some bits can select 2 bits, 3 bits, 4 bits and quantization bits. Are known.

Of course, in order to obtain good sound quality, the sampling frequency Fs needs to be 8 kHz and the number of quantization bits must be 4 bits (the bit rate at this time is 32 kbps). However, since the recording time is inversely proportional to the bit rate, it is often set to about 24 kbps in consideration of the sound quality.

[0005]

As described above, in the conventional recording apparatus, the sound quality is sacrificed when it is desired to record for a long time. However, it is not preferable to sacrifice sound quality, and it is desired to enable long-term recording while maintaining sufficient sound quality.

The present invention has been made in view of the above problems, and an object of the present invention is to provide a recording apparatus capable of recording for a long time while maintaining sound quality.

[0007]

SUMMARY OF THE INVENTION The present invention provides an audio signal A
In a recording device for converting digital data by an A / D converter and recording the digital data, a frequency characteristic judging means for judging a frequency characteristic of an input audio signal, and an A / D converter according to the judged frequency characteristic. And a sampling frequency changing means for changing the sampling frequency.

Further, according to the present invention, an audio signal input to a recording device which converts an audio signal into digital data by an A / D converter and then converts the audio data into coded data of a predetermined bit by an encoder for recording. The frequency characteristic determining means for determining the frequency characteristic of, the sampling frequency changing means for changing the sampling frequency in the A / D converter according to the determined frequency characteristic, and the encoding according to the determined frequency characteristic. And a bit number changing means for changing the bit number of the obtained code data.

Further, according to the present invention, the frequency characteristic determining means detects the repetition frequency of the voice signal waveform corresponding to the fundamental frequency of the vowel part of the voice, determines whether it is a low frequency voice or a high frequency voice, and The sampling frequency changing means is characterized by lowering the sampling frequency in the case of low frequency sound and increasing the sampling frequency in the case of high frequency sound.

Further, according to the present invention, the frequency characteristic judging means detects whether or not there are many components below a predetermined frequency, judges whether it is a low frequency sound or a high frequency sound, and the sampling frequency changing means. Is characterized by lowering the sampling frequency in the case of low frequency speech and increasing the sampling frequency in the case of high frequency speech.

[0011]

The present invention has been made by paying attention to the fact that male and female have different voice characteristics. In speech, vowels are identified by the peak position that forms the maximum value of the spectrum called Holtmant,
The ones with lower frequencies are called the first and second holts. The first and second holtmants are at 280 Hz to 2310 Hz for men and about 340 Hz to 2830 Hz for women. Here, Table 1 shows the positions of the Holtmants, which are the averages of the results of the vocalization measurement for each of 25 men and women.

[0012]

[Table 1] Also, the third holtmant is an important factor for speaker identification, which forms the characteristics of the individual's voice. Therefore, when recording voice, it is necessary to cover this third Holtmant. And this third
Holtmant is up to 3000Hz for men and 40 for women.
It is in the band up to 00 Hz. When performing A / D conversion while maintaining the sound quality of the audio signal, it is necessary to sample at a frequency twice that frequency. Therefore, as a sampling frequency for A / D conversion, 8 kHz is necessary for a female and 6 kHz is sufficient for a male.

As described above, the sampling frequency required for maintaining the sound quality differs between men and women. Therefore, by changing the sampling frequency between male and female, the efficiency of data compression can be increased while maintaining sufficient sound quality.

According to the present invention, the frequency characteristic of the input voice signal is determined. This distinguishes between male and female. Then, by changing the sampling frequency at the time of A / D conversion according to the identification result, efficient data compression can be performed without degrading the sound quality. For example, for men, the sampling frequency is 6 kHz,
In the case of a woman, by setting the frequency to 8 kHz, the data compression efficiency can be improved without losing the third Holtmant.

Further, if the digital data obtained by the A / D conversion is stored as it is, the data amount becomes enormous. Therefore, in the present invention, the data is stored after being encoded to reduce the data amount. Then, the number of bits of the encoded data obtained at the time of encoding is changed according to the frequency characteristic. For example, 4 bits for male voice,
If it is a female voice, set it to 3 bits. In particular, the sampling frequency for A / D conversion in this case is 6k for men.
Hz and 8 kHz for a female, 24 kbps for both male and female. In this way, it is possible to effectively compress data with the same bit rate.

Further, the waveform of the vowel portion of the voice is such that the same waveform is repeated at the fundamental frequency determined by the shape of the vocal cord and the like. And this fundamental frequency is different between men and women. For example, the fundamental frequencies of the voices of men and women have a distribution as shown in FIG. 4, with the distribution centering around 125 Hz for men and 240 Hz for women. Therefore, it is possible to determine male or female based on the fundamental frequency of the voice signal.

According to the present invention, the fundamental frequency is detected to determine whether the low frequency voice corresponding to the male voice and the high frequency voice corresponding to the female voice. Since the sampling frequency at the time of A / D conversion is changed according to this determination result, the data compression rate can be increased without impairing the sound quality.

When frequency analysis is performed on the voices of men and women,
Compared to male voices, female voices are characterized by having very few components below 150 Hz.

In the present invention, this predetermined frequency (for example, 1
It is determined by whether or not there are many components below 50 Hz), that is, a low frequency voice corresponding to a male voice and a high frequency voice corresponding to a female voice. Therefore, similarly to the case described above, the data compression rate can be increased without deteriorating the sound quality.

[0020]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the overall configuration of the system, including a solid-state recording / playback LSI 10 and an audio memory 3.
0 and the microcomputer 40. An analog input voice signal converted into an electric signal by a microphone or the like is input to the solid-state recording / reproducing LSI 10. The solid-state recording / reproducing LSI 10 includes therein an A / D converter 12, an encoder 14, a write driver 16, a read driver 1.
8, decoder 20, D / A converter 22, characteristic determination unit 24
And a control unit 26.

The input audio signal is input to the A / D converter 12, where it is converted into digital data of a predetermined bit, for example, 8 bits. The A / D converter 12 can switch the sampling frequency between two levels of at least 6 kHz and 8 kHz. The digital data from the A / D converter 12 is encoded by the encoder 14 by a predetermined method. For example, as the encoding method, MSBC
-AB and the like are preferable. This MSBC-AB means a modified sub-band coding that adaptively allocates bits, divides audio data into frames of about 20 msec, divides each frame into bands of about 4 bands, and divides each waveform. Encoding is performed, and the number of allocated bits for each frequency band at the time of encoding is adaptively changed.

The bit number (quantization bit number) of encoded data for one digital signal in the encoder 14 can be switched. The number of quantization bits can be switched to at least 3 bits or 4 bits.

The coded data thus obtained is
It is written in the audio memory 30 via the write driver 16. As a result, the input voice signal is recorded as digital data in the semiconductor voice memory 30. In addition,
The voice memory 30 is composed of, for example, a 4-Mbyte DRAM.

On the other hand, the encoded data read from the audio memory 30 via the read driver 18 is decoded by the decoder 20.
, Where it is decoded and converted into digital audio data. Then, the D / A converter 22 returns the analog output audio signal. Here, the decoding of the encoder 20 corresponds to the encoding of the encoder 14, the D / A converter 22 corresponds to the A / D converter 12, and these are recorded. The processing contents are changed in response to changes in the sampling frequency of the system and the number of quantization bits.

Then, in the present embodiment, the characteristic judging section 24 judges the characteristic of the input voice signal to judge whether it is a male voice or a female voice. In the case of a male voice, A for covering up to the third holtmant, which shows the features of the voice
The sampling frequency of the / D converter 12 may be 6 kHz. Therefore, when the characteristic determination unit 24 determines that the input audio signal is male, the control unit 26 causes the A / D converter 1 to operate.
Sampling frequency in 2 is 6 kHz, encoder 1
The quantization bits of 4 are set to 4 bits, and the decoder 20 and the A / D converter 22 are set to those corresponding to these.

On the other hand, when the characteristic judging section 24 judges that the input voice signal is female, the controlling section 26 sets the sampling frequency of the A / D converter 12 to 8 kHz and the quantization bit of the encoder 14 to 3 bits. Set to
The decoder 20 and the A / D converter 22 are set to correspond to these.

In this way, according to this embodiment, by changing the sampling frequency and the quantized bit while keeping the bit rate fixed at 24 kbps,
It is possible to reduce the amount of data stored in the audio memory 30 while maintaining the sound quality. That is, the data compression method can be changed according to the characteristics of the input audio signal, and effective data compression can be performed.

The microcomputer 40 controls the operation of the solid-state recording / reproducing LSI 10 in accordance with an operation signal input from the outside, and controls recording start / stop, reproduction start / stop, mode setting, and the like. The sampling frequency may be changed, the quantization bit may be changed, or the control based on the characteristic determination may be turned on / off by an external operation.

Next, an example of the configuration of the characteristic determining section 24 will be described with reference to FIG. In this example, the fundamental frequency of the input voice signal is measured to determine whether it is male or female. That is, the input audio signal is input to the low pass filter 52, where high frequency components are cut off, and then input to the frequency analyzer 52 for frequency analysis. And the determiner 5
6 determines whether the input voice signal is male or female, according to the result of the frequency analysis.

Here, in the low-pass filter 52, components of, for example, 500 Hz or higher are cut. A voice has various sounds by superposing a high-frequency component due to the shape of the tongue, jaw, etc. on the sound of the fundamental frequency determined by the vocal cords. However, the fundamental frequency of voice in a normal conversation is constant. For this reason,
If the component of 500 Hz or more is cut from the audio signal, only the component of the fundamental frequency remains. Therefore, the fundamental frequency can be detected by performing frequency analysis of the obtained signal.

The decision unit 56 is the frequency analyzer 54.
When the frequency is around 125 Hz according to the analysis result of 1., it is determined to be a male, and when the analysis result is around 245 Hz, it is determined to be a female. This determination may be made by comparing the count value of the counter with a predetermined value and determining whether it is within a predetermined range.

In this way, it is possible to determine whether the speaker is male or female by detecting the fundamental frequency of the voice. Therefore, by supplying this information to the control unit 26, the control unit 26 can compress the data in accordance with the characteristics of the audio signal.

Next, FIG. 3 shows another structural example of the characteristic judging section 24. In this example, it is determined whether the speaker is a male or a female according to the ratio of the component of 150 Hz or higher and the component of 150 Hz or lower of the audio signal. That is, the input audio signal is input to the low pass filter 62 and the high pass filter 64. The low pass filter 62 has 15
The signal of 0 Hz or higher is cut, and the high pass filter 64 cuts the signal of 150 Hz or lower.

The outputs of the low-pass filter 62 and the high-pass filter 64 are respectively level accumulators 66,
68 is input. These level accumulators 66 and 68 are
The signal level of the input signal is detected and this level value is integrated for a predetermined time. Therefore, the level accumulators 66 and 68 have a signal level of 150 Hz or less and
A signal level of 50 Hz or higher can be obtained. The signal of the integration result of the level integrators 66 and 68 is sent to the comparator 7
It is input to 0, and both are compared here.

In the case of a male, the signal level of the component below 150 Hz is large, while in the case of a female, the signal level of the component below 150 Hz is very small. Therefore, in the comparison result of the comparator 70, if the output of the level integrator 66, that is, the level of the signal of 150 Hz or less is larger, it is determined that the input voice signal is male, and the output of the level integrator 68, that is, If the level of the signal of 150 Hz or higher is higher, the input audio signal is judged to be female. In this way, the characteristic determining unit 24 of this example can determine whether the input audio signal is of a male type or a female type. Therefore, by using this circuit, data compression suitable for the characteristics of the audio signal can be performed in the same manner as in the above case.

Also, like the message recording of an answering machine,
In normal recording, the gender of the speaker is unknown. Therefore,
The determination of sex as described above must be performed early after the start of recording. Therefore, in the apparatus of the present embodiment, as shown in FIG. 5, for 10 seconds at the beginning of recording, the gender is determined while recording at a bit rate of 32 kbps with a sampling frequency of 8 Hz and 4 quantization bits. If the gender can be determined, the sampling frequency is changed to 6 kHz for men, and the quantization bit is set to 3 for women.
Change to bit and record at the bit rate of 24 kbps until the end. By doing this, although the sex determination time is a small percentage of the recording time, depending on the total recording time, the memory capacity required for one recording is close to the value at 24 kbps, which is sufficient. With a limited memory capacity, it is possible to record for a long time while maintaining excellent sound quality.

In the above-mentioned embodiment, the characteristic judging section 24 is described as receiving the analog input voice signal and performing the processing. However, the characteristic judging section 24 receives the digital data output from the A / D converter 12 and performs the processing. You can go. In this case, the circuit is formed entirely of digital circuits.

[0038]

As described above, according to the present invention,
The frequency characteristic of the input audio signal is determined. This makes it possible to distinguish between men and women. Therefore, by changing the sampling frequency at the time of A / D conversion according to this identification result, it is possible to suppress the loss of sound quality and perform efficient data compression.

Further, in the present invention, the data is stored after being encoded so as to reduce the amount of data, and the number of bits of the encoded data obtained by this encoding is changed according to the frequency characteristic. In this way, by controlling the number of quantization bits together, it is possible to effectively compress data while always keeping the same bit rate.

Further, in the present invention, this fundamental frequency is detected to determine whether the low frequency voice corresponding to the male voice and the high frequency voice corresponding to the female voice. Since the sampling frequency at the time of A / D conversion is changed according to this determination result, the data compression rate can be increased without impairing the sound quality.

Further, in the present invention, it is determined whether there is a low frequency voice corresponding to a male voice and a high frequency voice corresponding to a female voice depending on whether or not there are many components below the predetermined frequency. Therefore, similarly to the case described above, the data compression rate can be increased without deteriorating the sound quality.

Since such efficient data compression can be performed, recording can be performed for a long time with a limited memory capacity while maintaining sound quality.

[Brief description of drawings]

FIG. 1 is a block diagram showing an overall configuration of an embodiment.

FIG. 2 is a block diagram illustrating a configuration example of a characteristic determination unit.

FIG. 3 is a block diagram illustrating another configuration example of a characteristic determination unit.

FIG. 4 is a diagram showing a distribution of fundamental frequencies.

FIG. 5 is an explanatory diagram showing an operation at the time of recording.

[Explanation of symbols]

 10 Solid-state recording / playback LSI 12 A / D converter 14 Encoder 20 Decoder 22 D / A converter 30 Voice memory

Claims (4)

[Claims] 1. A recording device for converting an audio signal into digital data by an A / D converter and recording the digital data, and a frequency characteristic judging means for judging a frequency characteristic of an input audio signal, and a frequency characteristic judging means for judging the frequency characteristic. And a sampling frequency changing means for changing the sampling frequency in the A / D converter. 2. In a recording device which converts an audio signal into digital data by an A / D converter and then converts it into coded data of a predetermined bit by an encoder for recording, a frequency characteristic of an input audio signal is measured. Frequency characteristic judging means for judging, sampling frequency changing means for changing a sampling frequency in the A / D converter according to the judged frequency characteristic, and code data obtained by the coding according to the judged frequency characteristic. A recording device comprising: a bit number changing means for changing the bit number. 3. The recording device according to claim 1, wherein the frequency characteristic determining means detects a repetition frequency of a voice signal waveform corresponding to a fundamental frequency of a vowel part of voice, and selects either low frequency voice or high frequency voice. The recording device is characterized in that the sampling frequency changing means lowers the sampling frequency in the case of low frequency sound and increases the sampling frequency in the case of high frequency sound. 4. The recording device according to claim 1, wherein the frequency characteristic determination means detects whether there are many components having a frequency equal to or lower than a predetermined frequency and determines whether the component is a low frequency voice or a high frequency voice. The recording device characterized in that the sampling frequency changing means lowers the sampling frequency in the case of low-frequency sound and increases the sampling frequency in the case of high-frequency sound.