JPH04293100A - Digital sound recording and reproduction device using solid recording element - Google Patents

Abstract

PURPOSE:To offer the digital sound recording and reproduction device using a solid recording element which can perform sound recording and reproduction at a low bit rate. CONSTITUTION:An inputted voice is divided into a low frequency area and a high frequency area; and the energy level of the voice signal in the low frequency area is recorded in an energy level recording area R1 and the waveform of the voice signal in the low frequency area is recorded in low frequency recording areas D1-D24. Plural kinds of pattern voice signal are prepared for the high frequency area and when the voice signal is inputted, a pattern voice signal of frequency close to this voice signal is selected; while its code number is recorded in code recording areas R2 and R3, envelope data on the extracted voice signal is recorded in envelope data recording areas delta1-delta3 and epsilon1-epsilon2. When the reproduction is performed, they are read out respectively and put together to generate a voice signal by approximation.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital recording/reproducing apparatus using a solid-state recording element, and more particularly to a digital recording/reproducing apparatus using a solid-state recording element for recording audio signals.

0002

2. Description of the Related Art In recent years, recording and reproducing devices using solid-state recording elements (for example, voice synthesis LSIs) have been developed and are used in various devices. Unlike tape recorders, etc., this does not use magnetic tape or magnetic disks, so there is no need for a tape running mechanism or a motor.This makes the entire device smaller and lighter, and eliminates the occurrence of wow and flutter. It can be prevented. Furthermore, since there is no need to consider wear and tear of magnetic members such as tapes, it can be used semi-permanently and has the advantage of being highly reliable.

[0003] Therefore, for example, devices such as alarm devices and medical equipment that require voice commands to operate the surroundings without fail, and products such as vending machines and microwave ovens that have different characteristics. It is used in the required devices. Furthermore, due to its high reliability and ease of random access, its range of applications is expanding to include telephones and the like.

0004

[Problems to be Solved by the Invention] However, in recording and reproducing devices that use solid-state recording elements, audio is first digitized and stored, so when covering a wide range of frequencies, especially when the frequency is high, the memory capacity becomes insufficient. There was a problem that the capacity could not keep up and recording could only be done for a short period of time.

[0005] Therefore, in order to reduce memory consumption, it is necessary to compress the information of the audio signal and record it in memory, but the conventional methods are still insufficient, and it is necessary to compress the information of the audio signal and record it in the memory at a lower bit rate. It is important to encode while ensuring sound quality.

The present invention has been made in view of the above-mentioned problems, and its purpose is to provide a digital recording and reproducing device using a solid-state recording element that can obtain a predetermined sound quality at a practical bit rate. It is in.

[0007]

[Means for Solving the Problems] In order to solve the above-mentioned problems, a digital recording/playback device using a solid-state recording element according to the present invention has a low frequency range in which an audio signal can be converted into PCM, and a frequency dividing means for dividing the signal into the above-mentioned high frequency region; a PCM converting means for waveform encoding the signal in the low frequency region; an average frequency measuring means for measuring the average frequency of the signal in the high frequency region; Envelope information extraction means for extracting frequency domain envelope information; PCM information waveform encoded by the PCM conversion means; frequency information of the signal measured by the average frequency measurement means; extracted by the envelope information extraction means. and recording means for recording the envelope information.

Furthermore, in the digital recording/playback device using the solid-state recording element according to the present invention, the low frequency region of the audio signal is recorded as waveform-encoded PCM information, and the high frequency region is recorded as frequency information and envelope information. A digital playback device using a solid-state storage element that reproduces audio from a solid-state storage element in which audio is recorded includes a PCM information reading means for reading the PCM information from the storage element, and a low frequency region of the audio signal from the PCM information. a low frequency region creating means for creating a low frequency region; a frequency information reading means for reading the frequency information from the storage element; an envelope information reading means for reading the envelope information from the storage element; a codebook having a plurality of types of recorded pattern audio signals; a pattern audio signal reading means for reading a pattern audio signal corresponding to the frequency information from the codebook; and an envelope of the read pattern audio signal; a high frequency region creating means for creating a high frequency region of an audio signal by modifying it, a low frequency region of the audio signal created by the low frequency region creating means, and an audio signal created by the high frequency region creating means. and an audio signal creation means for creating an audio signal by synthesizing the high frequency region of the audio signal and the high frequency region of the audio signal.

[0009]

[Operation] In the digital recording/playback device using the solid-state recording element of the present invention having the above-described configuration, the audio signal, which is an important part constituting the basis of audio and has a low frequency, is divided by the frequency division means. PCM is possible because of the low frequency region and PCM is possible because of the high frequency.
It is divided into high frequency range, which is difficult to achieve. Among these, the audio signals in the low frequency range are converted into PCM by the PCM conversion means and then recorded by the recording means. On the other hand, the average frequency and envelope information of the audio signal in the high frequency region are simultaneously analyzed and recorded.

[0010] Correspondingly, during reproduction, an audio signal in the low frequency region is created from the PCM information read from the solid-state recording element, and this is synthesized with the high frequency region of the audio signal created separately. The audio signal is synthesized by At this time, when synthesizing the high frequency region, the pattern audio signal installed in the codebook is read by comparing it with the frequency information read from the solid-state recording element, and the envelope information at the time of recording of this audio signal is added to this. are superimposed to synthesize the high frequency region of the audio signal. In this way, the high frequency region of the audio signal is approximately synthesized, but the low frequency region, which is the essential part of the audio signal, is faithfully converted to PCM.
Relatively faithful recording and playback is possible.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below with reference to the drawings.

[0012] Due to the generation mechanism, audio signals can be divided into voiced sounds (100 to 2500 Hz) and unvoiced sounds (2500 to 500 Hz).
0Hz), and these can be handled separately by band division due to their different spectra. For this reason, in devices that perform recording and playback using solid-state recording elements, audio signals are usually handled after being appropriately divided into bands.

In this embodiment, the audio signal was band-divided into band 1, band 2, and band 3 as shown in FIG. Among these, band 1 (0 to 850 Hz) is an important part that forms the basis of audio. and band 2
(850 to 2500 Hz) does not have as much energy as band 1, but is a spectrum necessary for vowel identification. Furthermore, although band 3 (2500 to 5000 Hz) has less energy, it forms a band of unvoiced sounds, and is an important part for the generation of consonants including unvoiced sounds, such as "sashisu seso". Among these, band 1 (low frequency region) has a low frequency and can be sampled by PCM, but bands 2 and 3 (high frequency region) have high frequencies, so sampling by PCM and waveform encoding by quantization are possible. However, the bit efficiency is poor and it cannot be stored within the limited bit capacity.

Therefore, in this embodiment, the audio signal in the high frequency region (band 2 and band 3) is created by combining a pattern audio signal close to the input audio signal and envelope data (envelope information) of the input audio signal. By compositing, the bit efficiency is improved by creating an approximation.

That is, in the low frequency region (band 1), PCM waveform encoding is performed and recorded, but in the high frequency region (bands 2 and 3), the average frequency of the input audio signal and Envelope data is analyzed. Then, the analyzed envelope data is recorded, and the frequency of the pattern audio signal in the preset codebook is compared with the average frequency of the audio signal, and the frequency that is closest to the average frequency of the input audio signal is compared. A code of a pattern audio signal having a frequency is recorded as frequency information. Also, during playback, the code of this pattern audio signal is read, the pattern audio signal is called from the codebook, and the envelope data is superimposed on it to approximate the high frequency region (band 2 and band 3). creation is performed.

[0016] In this way, even if the high frequency region of the audio signal is created approximately, the low frequency region, which forms the important basis of audio, is recorded and reproduced with relative fidelity, so the sound quality is sufficiently guaranteed. It is possible to obtain an audio signal.

FIG. 2 is a format diagram showing the format of the recording element used in this embodiment. This format uses 3 words with 4 bits per word.
Two words are used and treated as one frame. Therefore,
The storage capacity per frame is 128 bits, and 1
When operating at a bit rate of 0 kb/sec, the duration of one frame is 12.8 msec. Also, since it is for solid-state recording, there is no special frame delimiter.
Identified by memory address.

In the figure, 24 words (4×24=96 bits) from D1 to D24 are an area (waveform recording area) where the audio waveform of band 1 is faithfully recorded.
1 (3 bits) is an area (energy level recording area) in which energy level information of the low frequency recording area is recorded. Furthermore, R2 and R3 are code recording areas in which codes of pattern audio signals installed in the codebook of the high frequency region are recorded, and δ1 to δ3
and ε1 to ε2 are envelope data recording areas in which envelope data in a high frequency range is recorded. Of these, the code of the pattern audio signal of band 2 is recorded in R2 (5 bits), and the code of the pattern audio signal of band 3 is recorded in R3 (4 bits). Then, envelope data of band 2 is recorded in δ1 to δ3, and envelope data of band 3 is recorded in ε1 to ε2.

FIG. 3 shows the recorded contents of the energy level recording area (R1; 3 bits) and the code recording area (R2; 5 bits and R3; 4 bits) in the format of this embodiment.

[0020] Here, the energy level recording area (R1) having a storage capacity of 3 bits can set eight energy levels. Therefore, when the energy level of the input signal is determined, it is determined to be one of eight energy levels between the highest level and the lowest level and is recorded.

Furthermore, R2 has a capacity of 5 bits.
In this embodiment, 22 of the 32 steps are used, and 22 types of codebooks are recorded. Furthermore, in band 3, 16 types of codebooks are recorded using all 4-bit storage capacity.

In this way, the pattern audio signal recorded in the code recording area has the band 2 frequency band (850 to 2500 Hz) divided into 22 stages, and the band 3 frequency band (2500 to 2500 Hz) divided into 22 stages. 5000H
z) is divided into 16 stages. Standard waveform audio signals at each divided frequency are recorded in the codebook as pattern audio signals, and can be easily handled from within the codebook by specifying the code numbers assigned to them. It is possible to extract patterned audio signals.

Furthermore, in this embodiment, the pattern audio signals recorded in advance in the codebook are selected based only on the average frequency in the high frequency region. and,
In the embodiment, this frequency is determined by the number of cycles of zero crossings in the signal. Therefore, when an audio signal is input, the number of zero-cross cycles of the audio signal is measured, the pattern audio signal having the closest cycle number is selected, and its code number is recorded in the code recording area. In addition, in R2, the first half (01 in Figure 3b)
010) are male voices, and the latter half (01011 in Figure 3b) are male voices.
(after) are set to female voices, while in R3, the first half (before 0111 in Figure 3c) is set to a male voice, and the second half (before 0111 in Figure 3c) is set to a female voice.
1000 onwards) are recorded with a female voice, which improves the audio fidelity during recording and playback, and can be expected to differentiate between male and female voices.

FIG. 4 shows how the information of the input audio signal is compressed when the audio signal in the low frequency recording area is recorded. That is, after the audio signal per word (per 0.4 msec) is modulated into 14-bit PCM, it is compressed into 7 bits quasi-instantaneously, taking into account the energy input at the same time. Then, the signal compressed quasi-instantaneously to 7 bits is compressed to 4 bits by non-linear compression and then recorded.
This is one word (4 bits) of the waveform recording area in the format shown in FIG.

Furthermore, as shown in FIG. 5, the level of energy recorded in R1 is not evenly divided, and as the energy increases, the range of energy divided becomes wider. It is possible to record and playback that follows the physiology of the body.

FIG. 6 is a block diagram showing the functional configuration of the recording circuit of this embodiment.

In the figure, first, the audio signal S is connected to the terminal 101.
1 is input, it is sampled at 3.84 MHz by the A/D converter 102 and further filtered by the first filter 103 to remove noise. The audio signal B generated in this manner is supplied to the second filter 104, the third filter 114, and the fourth filter 124.

At this time, the first filter 103
Since it is sampled in two stages using an 8-bit moving average, the audio signal B has a frequency of 30 kHz (3.84 MHz/128 = 30
kHz).

Note that the second filter 104 is 1875H.
It has a sampling frequency of z, and among the audio signals B, only the audio signal B1 belonging to band 1 is passed. Then, the third filter 114 passes only the audio signal B2 belonging to band 2 of the audio signal B, and further filters the fourth filter 114.
The filter 124 passes only the audio signal B3 belonging to band 3 of the audio signal B.

Among these, the audio signal B1 (low frequency region; band 1) passes through the second filter 104, and then
R1 is supplied to the determination circuit 105 and shift register 106. In the R1 determination circuit 105, the energy level of the audio signal B1 is detected, and R1 is determined as a digital value.
recorded in

On the other hand, the information of the audio signal B1 shown in FIG. 4 is compressed from the shift register 106 to the gate 107 and the compression conversion ROM 108. That is, a 14-bit/24-stage shift register 106
When the 14-bit information is input to the gate 107, the energy level input at the same time is taken into account and compressed into a 7-bit information signal quasi-instantaneously. This 7-bit information signal is then stored in a compression conversion ROM.
The data is further compressed into 4 bits by 108. In this way, the waveform information of the audio signal B1, which was 14 bits x 24 stages before being input to the shift register 106, is changed to 4 bits x 2 stages when recorded in the memory 110.
It is compressed and recorded into a four-stage information signal (D1 in Figure 2).
~D24) This is the case.

As mentioned above, in the low frequency region,
Since the level of the audio signal and the waveform of the audio signal are separated and recorded, the waveform of the audio signal is faithfully recorded, allowing relatively faithful recording and reproduction.

On the other hand, in the high frequency region (band 2 and band 3), the following approximation method is used. Unlike the low frequency region where the waveform of the input audio signal is input as is, the input signal is compared with a pattern audio signal set in a codebook 130 (not shown) prepared in advance, and The pattern audio signal having the frequency closest to the input signal will be selected and its code number will be recorded in memory 110.

The audio signal B2 that has passed through the third filter 114 is supplied to an R2 determination circuit 115 and an envelope detection circuit 116. Among these, R2 decision circuit 1
In step 15, it is determined which of the pattern audio signals set in the codebook 130 (not shown) the frequency of the input audio signal B2 is closest to. Then, when the closest pattern audio signal is selected, the code number of this pattern audio signal is transferred to the buffer register 120. The data is recorded in the memory 110 (R2 in FIG. 2). On the other hand, the envelope detection circuit 116 detects the envelope data of the input audio signal B2, and records it in the memory 110 via the buffer register 120 (δ1 to δ3 in FIG. 2).

Furthermore, the audio signal B3 that has passed through the fourth filter 124 is supplied to an R3 determination circuit 125 and an envelope detection circuit 126, in the same way as the audio signal B2, and the frequency information and envelope data are analyzed, respectively. , frequency information (code number of the pattern audio signal) and envelope data are stored in the buffer register 120.
(R3 and ε1 to ε2 in FIG. 2).

[0036] Here, the code recording areas (R2 and R3
) The audio modification to the waveform is performed using the envelope data. Note that δ1 to δ3 at this time
The recorded contents of and ε1 to ε2 in this embodiment are shown in FIG.

That is, as shown in FIG. 8a, sampling points for the audio signal of band 2 are obtained by dividing one frame into three equal parts. Then, the maximum absolute value of the data of 30 samples near this sampling point is determined, and the magnitude of the amplitude at the sampling point is determined.
2 is recorded.

Furthermore, since the envelope data of the band 3 audio signal does not require as much distinguishability as the envelope data of the band 2 audio signal, two-point sampling is performed as shown in FIG. 9a.

On the other hand, during playback, a high frequency domain audio signal is created by combining envelope data as shown in FIG. 8a and a pattern audio signal as shown in FIG. 8b. In the embodiment, the amplitude of envelope data is recorded in 16 steps of 3 dB each, and is treated as an envelope of an audio signal synthesized by linear interpolation between sampling points. Here, the number of samples subjected to linear interpolation is 63 points for the band 2 audio signal and 95 points for the band 3 audio signal.

In the audio signal B2 in the band 2 area, the pattern audio signal corresponding to the code number recorded in R2 is called out from the codebook, and at the same time, the envelope data is read out from δ1 to δ3. By combining these, it is possible to create an audio signal that approximates the input audio signal B2.

In the audio signal B3 in the band 3 area, there is a pattern audio signal corresponding to the code number recorded in R3. At the same time as being called from the codebook, envelope data is read from ε1 to ε2. And by combining these,
It is possible to create an audio signal similar to the input audio signal B3.

Furthermore, the audio signal B1 is extracted from the low frequency recording area.
The information regarding is read and shown in FIG.
a (B1; band 1) and Fig. 10b (B2; band 2
) and FIG. 10c (B3; band 3) are superimposed, it is possible to create an audio signal similar to the input audio signal B. As a result, input terminal 10
Audio signal S1 approximated to audio signal S1 input from 1
It becomes possible to output 2. The audio signal created in this way is obtained by faithfully recording and reproducing the band 1 signal, which is the important basis for audio signal identification, and by approximating and recording and reproducing the band 2 and band 3 signals, which have less contribution. Because of this, we are able to create smooth synthesized speech that allows the content of the speech to be fully understood.

[0043] The codebook data is sampled at 8 bits/15kHz, and the waveform of the pattern audio signal is adjusted so that the start and end rises cross zero at the zero point so that it continues smoothly. ing.

FIG. 11 is a block diagram showing the functional configuration of the reproducing apparatus of this embodiment. First, when the energy level information of R1 is read from the memory 110 by the R1 register 151, the waveform recording area (D1 to D24) is simultaneously read by the band 1 data register 152.
data is read. Then, the 4-bit/1-word information read by the band 1 data register 152 is expanded into 7-bit information by the data expansion ROM 153. Then, this expanded signal is interpolated in an interpolation circuit 154 based on the signal from the R1 register 151, and a 15kHz/14-bit low frequency audio signal C is generated.
1 (corresponding to the input phonetic symbol B1), which is input to the adder 155. Further, when R2 is read from the code recording area by the R2 register 161, the corresponding pattern audio signal P1 is selected from the codebook ROM 162.

When this selected pattern audio signal P1 is input to the adder 163, the R2 register 161
Envelope data δ1 to δ3 read by
is read by the envelope data register 164, interpolated into 15 kHz/8-bit information by the interpolation circuit 165, and then input to the multiplier 163. In this multiplier 163, the envelope data is superimposed on the pattern audio signal P1, and the input audio signal B
An audio signal C2 similar to 2 is created and input to the adder 155.

Furthermore, R recorded in the code recording area
3 is read by the R3 register 171, this is referenced with the pattern audio signal in the codebook 172, and when the pattern audio signal P2 is selected, it is input to the multiplier 173. On the other hand, the envelope data ε1 to ε2 read simultaneously at this time are input to the envelope data register 174, and the interpolation circuit 175
The signal is interpolated into 15kHz/8-bit information and input to the multiplier 173. Then, the envelope data of the pattern audio signal P2 is superimposed to create an audio signal C3 that approximates the input audio signal B3, and this is input to the adder 155.

In this way, when the low frequency domain audio signal and the high frequency domain audio signal are respectively input to the adder 155, these are synthesized in the adder 155, and 15 kHz, 14 bit information is generated. Mama, D
The signal is outputted from the output 180 via the /A converter 180 as an audio signal S2.

Comparing the input audio signal S1 and the output audio signal S2 at this time, it is found that although the low frequency region is faithfully recorded and played back, the high frequency region is synthesized by approximation. , the fidelity to the original sound cannot be expected, and although it is possible to distinguish between female and male voices, it is not clear whose voice it is. However, you can hear smooth audio with relatively little noise and distortion, and you can fully understand what is being said.

To illustrate this specifically, the dynamic range is 66 dB in band 1, and 9 dB in bands 2 and 3.
It was 3 dB, and the total was 65 dB. Furthermore, the frequency characteristics are flat from DC to 5kHz, which is ±2d.
It was ranked B. Furthermore, the distortion rate was approximately 10% below 850 Hz (however, the value at 450 Hz).

[0050]

[Effects of the Invention] In the digital recording and reproducing apparatus using the solid state recording element of the present invention constructed as described above,
Audio information can be compressed using simple hardware, making it possible to perform practical recording and playback at a low bit rate.

[Brief explanation of drawings]

FIG. 1 is an explanatory diagram showing frequency bands divided into bands in an embodiment.

FIG. 2 is a format diagram showing the contents of the format used in the memory of this embodiment.

FIG. 3 is an explanatory diagram illustrating the recorded contents of a format.

FIG. 4 is an explanatory diagram illustrating bit compression.

FIG. 5 is an explanatory diagram illustrating the recorded contents of an energy level recording area and a code recording area.

FIG. 6 is a block diagram showing the functional configuration of the recording device of this embodiment.

FIG. 7 is an explanatory diagram illustrating the recorded contents of an envelope data recording area.

FIG. 8 is an explanatory diagram illustrating how a band 2 audio signal is created.

FIG. 9 is an explanatory diagram illustrating how band 3 audio signals are synthesized.

FIG. 10 is an explanatory diagram illustrating how an audio signal is created.

FIG. 11 is a block diagram showing the functional configuration of the reproducing circuit of this embodiment.

[Explanation of symbols]

R1 Energy level recording area R2, R3 code recording area D1 to D24 Waveform recording area

Claims (2)

[Claims] 1. Frequency dividing means for dividing an audio signal into a low frequency region that can be converted into PCM and a higher frequency region, PCM converting means for waveform encoding the signal in the low frequency region, and average frequency measuring means for measuring the average frequency of the signal in the frequency domain; envelope information extraction means for extracting envelope information in the high frequency domain;
Recording means for recording PCM information waveform encoded by the PCM conversion means, frequency information of the signal measured by the average frequency measurement means, and envelope information extracted by the envelope information extraction means;
A digital recording device using a solid-state recording element, characterized by comprising: 2. A solid-state recording element that reproduces audio from a solid-state storage element in which a low frequency region of an audio signal is recorded as waveform-encoded PCM information, and a high frequency region is recorded as frequency information and envelope information. In the digital playback device used, PCM information reading means for reading the PCM information from the storage element, low frequency region creation means for creating a low frequency region of an audio signal from the PCM information,
frequency information reading means for reading the frequency information from the storage element; envelope information reading means for reading the envelope information from the storage element; and a plurality of types of patterned audio signals recorded for each frequency as patterned audio signals. a codebook having a codebook; a pattern audio signal reading means for reading a pattern audio signal corresponding to the frequency information from the codebook; A high frequency region creating means for creating a region, a low frequency region of the audio signal created by the low frequency region creating means, and a high frequency region of the audio signal created by the high frequency region creating means. 1. A digital playback device using a solid-state recording element, characterized in that it includes: audio signal creation means for creating an audio signal.