JPS58215696A - Real time voice coder - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION (1) Technical field of the invention The present invention relates to a real-time speech encoding device, and more specifically,
An LSI that encodes and synthesizes audio data in real time (
The present invention relates to a real-time speech encoding device that uses a large-scale integrated circuit to determine the start and end of speech encoding.

(2) Background of the invention In recent years, various speech synthesis LSIs have been developed.

As speech synthesis methods, PAItCORADPCM method, phoneme one-sided method, etc. are known. In the PACOR method and the phoneme unilateral method, very complicated division is required to analyze speech, which takes too much time and money, so at present LSIs that only perform synthesis have been announced. , no LSI that performs real-time encoding is known. In this case, encoding is performed using hardware external to the LSI.

On the other hand, in the ADPCM method (adaptive differential PCM method), audio analysis and audio data compression are relatively simple, so audio can be encoded in a short time and at a low cost.
In addition, since the quality of the sound synthesized after encoding is close to that of natural sounds, real-time speech coding and synthesis LS
Examples implemented using the I'tADPcM method are known.The present invention relates to an apparatus for encoding audio data in real time and storing it in a memory, as in the above-mentioned ADPCM method.

(3) Conventional technology and problems Conventional real-time audio encoding/synthesis LSIs cannot process audio data input from a microphone or tape recorder, etc.
It has the function of sampling digital data converted by the eA-D converter, compressing and encoding the digital data, and decoding (synthesizing) audio digital data from the compressed and omitted data. It also has a function to play audio data. The compressed and encoded data is stored in memory.

Input audio data usually includes repeating sound periods and silent periods. For example, the period before and after the pronunciation of "good morning" is a silent period, and the period during which it is being pronounced is a sound period.

Previously, all sampled data was encoded and stored in memory, regardless of whether it was a sound period or a silent period1.
, Therefore, the data of the silent period is also stored in the memory,
There was a problem with inefficient memory usage.

(4) Purpose of the Invention In view of the problems in the prior art described above, the purpose of the present invention is to compress, encode, and store in a memory only the data of the sound period of input audio data in a real-time audio encoding device. Based on this concept, the aim is to make effective use of memory or reduce the required memory capacity.

(5) Structure of the Invention Summary of the present invention for achieving the above objects: An A-D converter for converting input audio data into a digital signal;
- a sound period detection unit that samples the digital signal output from the D converter to detect a sound period of input audio data; starts encoding the digital signal from the start of the sound period; An encoder that finishes encoding a digital signal after a predetermined time from the end of a sound period or at the end of encoding a digital signal having a predetermined number of encoded bits, and encoded data output from the encoder. The real-time audio encoding device includes a memory IJ Th for counting the write address and outputting the write address, and a memory IJ Th for storing the encoded data according to the write address.

(6) Examples of the invention Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block circuit diagram showing a real-time speech encoding/synthesizing device according to an embodiment of the present invention, and FIG. 2 is a block circuit diagram illustrating a real-time speech encoding/synthesizing device according to an embodiment of the present invention. FIG. , is a waveform diagram showing input audio. Second
As shown in the figure, there is a silent period before and after the sound period between times t1 and t. The audio analog signal output from the microphone 1 has its high frequency cut by a low-pass filter 2 and is input to an A-D converter 3 . The operation of the A-D converter 3 is controlled by a control section 4.

That is, when a start signal St- is given to the control section 4 at time to (FIG. 2), an A-D conversion start signal a is given from the control section 4 to the A-D converter 3 in response to this, and thereby, The A/D converter 3 converts the analog signal from the low-pass filter 2 into, for example, an 8-bit digital signal over the silent period and the sound period. The 8-bit digital signal from the A/D converter 3 is passed through the data path 5 to the A/D converter 3.
The signal is input to a PCM encoder 6 and a sound period detector 7.

The sound period detection unit 7 samples the input 8-bit digital signal, detects the sound period of the input audio data from the sampling result, and outputs an encoding start signal at the start of the sound period. At the end of the encoding process, an encoding end signal e k Al) is given to the PCM encoder 6 . ADP
When the CM encoder 6 receives the encoding start signal b'l, the CM encoder 6 receives the encoding start signal b'l.
A bit digital signal is compressed and encoded into, for example, 4-bit data using the well-known ADPCM method. ADPC
For example, when compressing data using the M method, two consecutive 8-bit data are compressed into 01113100 and 01001100.
Then, the difference is 00110000, but if we take the upper 4 bits from the lower 3 bits of this numerical value, compressed 4-bit data 1100 is obtained. The compressed and encoded 4-bit data is input to the address counter 9 and the audio memory 10 via the path 8fc.

Address counter 9 counts input 4-bit data and provides a write address to memory 10 via path 11. 4-bit data is written into the memory 10 according to the write address.

Thus, according to the apparatus shown in FIG. 1, the input speech is encoded in real time, and substantially only the data of the sound period 1141 is stored in the memory.

Note that the data stored in the memory 10 is processed by the ADPCM decoder 12, the DA converter 13, and the low-pass filter 14.
Since it is possible that a silent period momentarily exists even during a sound period that is output as a synchronized voice to the speaker 15 via Do it using one of the traditional methods.

According to the first method, the total number of encoding bits used for compression encoding of audio data is specified in advance, and the counter 9
When the total number of encoded bits is counted, the voice encoding end signal de counter 9 outputs ADP(,'M n encoder 6
to end the encoding.

According to the second method, by utilizing the fact that a silent state usually continues for a certain period of time or more after the end of speech, the voiced period detecting section detects a silent period section for a certain period of time or more, and then generates an encoding end signal. c k Provided to the ADPCM encoder 7.

Note that the control unit 4, ADPCM encoder 6, and
The sound period detecting section 7 and the ADPCM decoder 12 are realized by one real-time speech encoding/synthesizing LSI.

In the embodiments described above, the case where 8-pit data is converted to 4-pit compressed encoded data for real-time audio encoding using the ADPCM method is used as an example, but the present invention is not limited to this, and the present invention is not limited to this. Effects of the InventionAs is clear from the above explanation, according to the present invention, real-time audio In the encoding device, only the sound period data of the input audio data is compressed and encoded and stored in the memory, so that the silent period data is not stored in the memory, resulting in a significant saving in memory capacity and money. can.

[Brief explanation of drawings]

FIG. 1 is a block circuit diagram showing a real-time speech encoding/synthesizing apparatus according to an embodiment of the present invention, and FIG. 2 is a waveform diagram showing an example of input speech input to the microphone 1 of FIG. 1. 1...Microphone, 2...Low pass filter, 3.
...A.■) Converter, 4.-.Control unit, 6..ADPC
M encoder, 7...Speech period detection section, 9...Address counter, 10...Audio memory. Patent applicant Fujitsu Ltd. Patent attorney Akira Yoiki Patent attorney Kazuyuki Nishidate (1) Yukio Patent attorney Akira Yamaguchi

Claims (1)

[Claims] 1. An A-D converter that converts input audio data into a gold digital signal; a sound period detection unit that samples the digital signal output from the A-D converter to detect a sound period of the input audio data; The encoding of the digital signal is started from the start of the sound period, and the digital signal is encoded after a predetermined time from the end of the sound period or at the end of encoding the digital signal with a predetermined number of encoded bits. an encoder that completes the encoding of the encoder, a counter that outputs the number of sets of encoded data output from the encoder and outputs a write address, and a memory IJ that stores the encoded data according to the write address.
A real-time speech encoding device comprising: