JPH02156298A - System and device for silent compression of voice - Google Patents

Abstract

PURPOSE:To eliminate the need of a buffer for storing encoded data by a 1-frame portion, and to contrive the reduction of a hardware by providing a register for storing temporarily head addresses of each frame. CONSTITUTION:An encoder 2 writes a digital voice waveform inputted to a terminal 1 in an encoded memory at every sample. On the other hand, a silent deciding device 12 decides whether a frame is a silent frame or a voiced frame at every one frame. A register 11 is a register for storing temporarily the same address as an address of the memory 8 indicated by a counter 7 in the end of each frame. That is, an address (a) of the memory 8 in which the head data of each frame is to be written is stored. In such a way, by writing directly encoded data in the memory 8 and executing easily rewriting to silent information data by an address management of the memory 8, a data buffer for storing a 1-frame portion becomes unnecessary and a hardware can be reduced.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a recording and reproducing apparatus using a voice waveform encoding method, and particularly to a silence compression method and apparatus.

[Conventional technology]

Digital audio waveform encoding methods can be classified into waveform encoding methods and source encoding methods. Waveform encoding methods include PCM, ADPCM, and DPCM.

There are ADM methods, etc., and source coding methods include LPC
, PARCOR, and LSP methods.

Generally, a waveform encoding method is often used because of its good sound quality. The waveform encoding method compresses each sample in the amplitude direction. However, since the bit rate is high in sample-by-sample processing, silence compression has come to be used to lower the bit rate.

The audio signal includes silent sections for breath breaks and the like.

Although this silent section differs depending on the individual and the state of vocalization, it corresponds to about 10 to 30% of the total voice section and accounts for a large proportion of the whole. Therefore, silence compression replaces the silent section included in the audio signal with data of a predetermined time. For example, tens to hundreds of samples can be used as one frame, and information in one frame of a silent section can be replaced with one word of data. Further, by outputting silence data for one frame based on the data compressed in this manner, it is possible to reproduce the original silent section.

Here, methods for determining whether it is a voice section or a silent section include, for example, a method of comparing the average value of samples of one frame with a predetermined threshold value, or a method of comparing all samples in one frame and a predetermined threshold value. A method of making a determination by comparing it with a predetermined threshold value is well known.

Next, the encoding of silence compression will be explained using a specific numerical example. When a digital audio waveform is encoded with 8 bits, levels from -128 to +127 can be expressed using two's complement representation.

Here, if the positive and negative levels are made equal during encoding, -12
The code representing 8 can be used as silence information data. That is, each time a silent frame appears, the entire frame can be replaced with one word of -128.

FIG. 3 shows an audio recording/playback device including a conventional silence compression circuit. 1 is a digital audio waveform input terminal, 2 is an encoder, 3 is a data buffer for storing encoded data for l frames, 4 is a silence determiner, and 5 is a memory in which silence information data indicating a silent frame is written. 6 is a write controller, 8 is a memory, 7 is a counter pointing to the address of memory 8, 9 is a decoder, 10 is a composite data output terminal, 13 is a silence information data detector, 14 is a silence data 16 is an output controller.

The operation of this device will be explained. The encoder 2 encodes the digital audio waveform input to the terminal l sample by sample and stores it in the data buffer 3. On the other hand, the silence determiner 4 determines whether each frame is a silent frame or a sound frame. If it is determined to be a sound frame, the data buffer 3
data is transferred to the memory 8. Furthermore, when it is determined that the frame is a silent frame, the data in the data buffer 3 is not transferred and the silent information data in the register 5 is written into the memory 8.

For example, assume that the Nth frame is a sound frame and the N+1th frame is a silent frame. Counter 7 always points to the address of memory 8 where the next data will be written.
Increment by 1 each time one word is written. The encoder 2 encodes the digital audio waveform of the Nth frame sample by sample, and stores the encoded data in the data buffer 3. After all the digital audio waveforms of the Nth frame are encoded and stored in the data buffer 3, the silence determiner 4 determines whether there is a voice or not. Since the Nth frame is a sound frame, the write controller 6 selects the data buffer 3 and transfers one frame worth of encoded data in the data buffer 3 to the memory 8.

At this time, the counter l is incremented by 1 every time l words are written. Next, as in the case of the Nth frame, the encoder 2 encodes the digital audio waveform of the N+1th frame and stores the encoded data in the data buffer 3. No. N
After all +1 frame of digital audio waveforms are encoded and stored in the data buffer 3, the silence determiner 4 determines whether there is a sound or not.
Since the N+1-th frame for which silence is determined is a silent frame, the writing controller 6 selects the register 5 and writes one word of silence information data (-128) into the memory 8. At this time, the counter 7 is incremented by one.

On the other hand, the silence information data detector 13 determines whether the encoded data read from the memory 8 is silence information data. If the data is other than silence information data, the decoder 9 decodes it and outputs composite data from the output terminal 10. If it is silent information data, stop the signal of counter 7,
The output controller 16 selects the register 14 and outputs silent data (0) with a time length of one frame from the output terminal 10.

[Problem to be solved by the invention]

However, the above-described conventional silence compression method requires a data buffer 3 for at least one frame, which has the disadvantage of increasing the amount of hardware. Furthermore, after processing the last data of each frame and determining whether there is a sound or no sound, if it is a sound section, all data in the data buffer 3 must be transferred to the memory 8, so it takes time to transfer. It takes a lot of time and the process is complicated.

[Means to solve the problem]

The speech silence compression method and device includes means for analyzing a digital speech waveform using a waveform encoding method, means for storing data encoded by the analysis means, and a speech silence compression method and apparatus for each predetermined number of samples. a determining means for determining silence; a rewriting means for rewriting sign-ratio data on the storage means into silence information data according to a determination result by the determining means; and completion of the voice/silence determination by the determining means. and means for returning the address on the storage means that is currently pointing to an address on the storage means that stores silence information data.

In this way, in the present invention, by temporarily storing the start address of each frame, it is possible to eliminate the need for a buffer for storing encoded data for one frame.

〔Example〕

Next, the present invention will be explained with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the present invention.

Components in FIG. 1 having the same numbers as those explained in FIG. 3 have the same functions as in FIG. 3. In addition, 11 is a register for temporarily storing the address of memory 8;
is a silence detector.

FIG. 4 is a waveform diagram showing sample points for two frames. Note that, to simplify the explanation, one frame is assumed to be 10 data here. The Nth frame is a sound frame, the N+1th frame is a silent frame, and the sample data of the MN frame is 81°S2. ..., SIO, No. N+1
The frame sample data is Sll, S12. ...,
Let it be S20. The data encoded by encoder 2 is CI
,..., CIO.

C11, . . . , C20.

In order to explain the writing of encoded data into the memory 8, FIG. 5 shows the memory 8 portion of FIG. 1 extracted and shown in detail.

The operation of this device will be explained. The encoder 2 encodes the digital audio waveform input to the terminal l sample by sample and writes it into the memory 8. On the other hand, the silence determiner 12 determines whether each frame is a silent frame or a sound frame.

The register 11 is a register that stores the same address as the address in the memory 8 pointed to by the counter 7 at the end of each frame. That is, it stores the address of the memory 8 where the first data of each frame is to be written. In this case, the register 11 stores address a of the memory 8 (FIG. 5(1)).

First, the encoder 2 encodes Sl of the Nth frame and writes encoded data C1 to address a of the memory 8. Increment counter 7 for each sample (Figure 5 (2))
, S2. ..., similar processing is performed for S10.

After encoding the last data S10 of the Nth frame and writing the encoded data C10 into the memory 8, the silence determiner 1
2 determines whether there is a sound or no sound. Since the Nth frame is a sound frame, the address a+lo is stored in the register 11 (FIG. 5(3)), and the process moves on to the N+1th frame.

The same processing as for the data of the Nth frame is performed for S11°..., S20 (FIG. 5 (4)). After writing the encoded data C20 into the memory 8, the silence determiner 12 determines whether there is a sound or not. Since the N+1st frame is a silent frame, the address a+lO stored in register 11
is returned to the counter 7 (Fig. 5 (5)), and the silence information data in the register 5 is written to address a+IO of the memory 8 (
Figure 5 (6)). The register 11 stores the next address a+11 where the silence information data is written (see FIG. 5).
7)).

FIG. 2 is a block diagram of another embodiment of the invention.

In FIG. 2, components having the same numbers as those explained in FIG. 3 have the same functions as in FIG. Additionally, 15 is a controller that controls the address counter 7 of the memory 8.

FIG. 6 shows a second portion extracted in order to explain the writing of encoded data into the memory 8 and the control of the counter 7 of the controller 15.

The operation of this embodiment will be explained. In this embodiment, since the counter 7 is controlled by the controller 15, the register 11 for temporarily storing the start address of each frame is not required.

The encoder 2 encodes the digital audio waveform input to the terminal 1 sample by sample and writes it into the memory 8. On the other hand, the silence determiner 4 determines whether each frame is a silent frame or a sound frame.

First, the encoder 2 encodes the Nth frame Sl and writes encoded data CI to address a of the memory 8. Increment counter 7 for each sample, S2. ...,
Similar processing is performed for SIO (Figure 6 (1) (2)
)). encode the last data SIO of the Nth frame,
After writing the encoded data CIO into the memory 8, the silence determiner 4 performs voice/silence determination. Since the N-th frame is a sound frame, processing moves to the N+1-th frame next. 8
11. ...For S20, the same processing as for the data of the Nth frame is performed (FIG. 6 (3)). Encoded data C2
After writing 0 into the memory 8, the silence determiner 4 performs voice/silence determination. Since the N+1 frame is a silent frame, the controller 15 subtracts 10 from the value of the counter 7 and returns the address one frame earlier (FIG. 6(4)), and transfers the silent information data in the register 5 to the memory 8. Write to address a+10 (Fig. 6 (5)).

The counter 7 is incremented ((6) in FIG. 7), and processing moves on to the next frame.

On the other hand, the silence information data detector 13 determines whether the encoded data read from the memory 8 is silence information data, and if the data is other than silence information data, the decoder 9
The decoded data is decoded and the combined data is output from the output terminal 10.

If it is silent information data, the counter 7 stops advancing, the output controller 16 selects the register 14, and outputs silent data (0) with a time length of 1 frame from the output terminal 10.

〔Effect of the invention〕

As described above, according to the present invention, encoded data is written directly into memory and rewritten to silence information data easily by memory address management, thereby eliminating the need for a data buffer and reducing hardware. Furthermore, since there is no need to transfer encoded data from the buffer to the memory, processing time is shortened.

Note that in the embodiment, the memory address was returned to the previous one by providing a register that stores the start address of each frame and a controller that calculates the memory address to which silent information data is written. It is clear that the same effect can be obtained in the silence compression method and apparatus of the present invention even if the address is moved back.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of a recording device including a silence compression circuit according to the present invention, FIG. 2 is a block diagram of an embodiment of a recording/playback device including a silence compression circuit according to the present invention, and FIG. 3 is a conventional block diagram. FIG. 4 is a waveform diagram showing the basic principle of the present invention, and FIGS. 5 and 6 show details of writing data to memory. FIG. 1...Digital audio waveform input terminal, 2...
...Encoder, 3...Data buffer, 4.
... Silence determiner, register in which silence information data is written, 6... Write controller, 7... Counter, 8... ...Memory, 9...Decoder, 10...Synthetic data output terminal, 11...Register for temporarily storing memory address, 12... ...silence detector,
13... Silence information data detector, 14...
...Register 5...Controller, 16...Output controller.

Claims (1)

[Scope of Claims] 1. In a speech compression device having an analysis means for analyzing a digital speech waveform using a waveform encoding method, a storage means for storing data encoded by the analysis means;
a counter that indicates an address of the storage means; a register that temporarily stores an address on the storage means that stores leading data of a predetermined number of samples;
determining means for determining whether there is a sound or no sound for each of the predetermined number of samples; and a code on the storage means that returns the address saved in the register to the counter according to the determination result by the determining means. 1. An audio silence compression device, comprising: rewriting means for rewriting encoded data into silence information data. 2. In a voice compression device having an analysis means for analyzing a digital voice waveform using a waveform encoding method, a storage means for storing data encoded by the analysis means;
a counter that indicates the address of the storage means; a determination means that determines whether there is a sound or no sound for each predetermined number of samples; and a controller that returns the address in the counter by the predetermined number of samples; and a rewriting means for rewriting encoded data on an address returned by the controller in the storage means into silence information data according to a determination result by the determination means. Device. 3. In a voice compression device having an analysis means for analyzing a digital voice waveform using a waveform encoding method, a storage means for storing data encoded by the analysis means;
an instruction means for specifying an address of the storage means; a determination means for determining whether or not there is a sound for each predetermined number of samples; and an address that returns the address in the instruction means by the predetermined number of samples. The present invention is characterized by comprising a converting means and a rewriting means for rewriting the encoded data on the address returned by the address converting means in the storage means into silence information data according to the determination result by the determining means. Audio silence compression method.