JP2602641B2 - Audio coding method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Overview] A frequency characteristic unique to a digital audio signal is extracted by a spectrum envelope information extraction circuit, and sound source information such as amplitude, bitch cycle, and distinction between sound / no sound is extracted by a sound source information extraction circuit. From the extracted information, the silence detection circuit detects the number of continuous silences in the analysis section of the audio signal, determines whether or not silence can be compressed based on the number of continuous silences, and converts the continuous silence data into silence information and continuous number information. Convert, compress the silence section,
It performs voice coding.

[Industrial applications]

The present invention relates to a voice coding method in which a voice signal converted into a digital signal is coded, and a silent section is further compressed to reduce the amount of stored information.

The voice response system used in various systems includes an editing system for playing back and editing recorded analog voice signals or digitized voice signals such as PCM signals, and analyzing and converting voice signals into parameters. An analysis synthesis method or the like in which the storage capacity is reduced and the storage capacity is reduced is known. LSP (line spectrum pai) as one of the analytical synthesis methods
The r) method analyzes a voice signal into frequency domain parameters, while the PARCOR method that analyzes time domain parameters is also known.

[Conventional technology]

FIG. 5 shows a block diagram of a conventional example for LSP analysis.
1 is an input terminal of an analog audio signal, 2 is an output terminal of LSP data, 3 'is an LSP analysis signal converting device, 4 is an AD converter for converting an analog audio signal into a digital audio signal, 5
Is a common bus, and 6 is a processor (CP
U) and 7 are memories (ME) for storing programs, data, etc.
M), 8 is a spectrum envelope information extraction unit, 9 is a storage unit, 10
Is a sound source information extraction unit, 11 is an amplitude extraction unit, 12 is a pitch period extraction unit, 13 is a sound / non-sound determination unit, 14 is a storage unit, and 15 is an output editing unit that edits and outputs analysis information.

The analog audio signal applied to the input terminal 1 is AD
The converted digital audio signal is converted into a digital signal by the conversion unit 4 and transferred to the spectrum envelope information extraction unit 8 and the sound source information extraction unit 10 directly or via the memory 7 under the control of the processor 6, and the sampling period ( By setting a frame period) and an analysis threshold value (silence determination value), spectrum envelope information (LSP parameter) as a frequency characteristic unique to the input audio signal is extracted in the spectrum envelope information extraction unit 8, and the sound source information extraction unit 1
The amplitude extraction unit 11 extracts the amplitude, the pitch period extraction unit 12 extracts the pitch period, and the sound / non-speech determining unit 13 determines the presence / absence of sound / no-sound. Storage units 9, 14
Is for storing the set values and data in the extraction process.

The output editing unit 15 edits and outputs the spectrum envelope information extracted by the spectrum envelope information extraction unit 8 and the sound source information such as the amplitude, the pitch period, and the sound / no-sound discrimination extracted by the sound source information extraction unit 10. Yes and sound / silence determination unit 13
For this reason, a silent parameter is forcibly inserted into an analysis section determined to be silent due to being equal to or less than the analysis threshold value.

FIG. 6 is an explanatory diagram of the LSP data of the analysis result, showing a case of a 6-byte configuration. In the figure, SB is a stop bit, T is a pitch period, T1 and T0 are frame periods (sampling periods) indicating the types of frame periods encoded, A is an amplitude, and ω1 to ω8 are LSP parameters. The frame period is selected based on the length of the sound piece and the like. For example, any one of three types of periods can be designated by “00” to “10”. Output editing section
At 15, the analysis result is edited and output, and is transferred from the output terminal 2 to a speech code storage device (not shown) and stored.

[Problems to be solved by the invention]

For example, in LSP analysis in conventional speech coding,
The silent part of the sound part included in the audio information is also analyzed as it is and stored as the LSP data. In normal voice information, silent sections are often continuous. Even in this case, the analysis result is output and stored for each continuous silent section.
Digitizing and storing voice information requires a considerable storage capacity, which is disadvantageous in that it is wasteful.

The present invention compresses the silent section as described above,
The purpose is to improve the storage efficiency.

[Means for solving the problem]

Referring to FIG. 1, the audio encoding system of the present invention converts an analog audio signal into a digital audio signal by an A / D converter 4 and the like, and converts the digital audio signal into a digital audio signal. A spectrum envelope information extraction unit 8 for extracting spectrum envelope information indicating a frequency characteristic inherent to a voice signal, and a voice signal converted into a digital signal,
A sound source information extracting unit 10 including an amplitude extracting unit 11, a pitch period extracting unit 12, a sound / silence determining unit 13 for extracting sound source information such as an amplitude, a pitch period of a voice signal, and distinction between a sound / no sound and a spectrum; The number of continuous silence frames is detected from the envelope information and the sound source information, and when the number of detected consecutive silence frames exceeds the number of reference frames for a sound section at the time of decoding, it is determined that silence compression is possible. A compression processing unit within a dashed line including a silence detection unit 16 that performs processing, a silence compression availability determination unit 17, a silence compression unit 18, and the like. The data of the other continuous silence frames are converted into the silence frame continuation number information, leaving the reference frames for the speech interval at the time of decoding in the number, and the silence interval included in the audio information is compressed. . Reference numeral 15 denotes an output editing unit.

(Operation)

The spectrum envelope information extraction unit 8 extracts spectrum envelope information from the audio signal converted into the digital signal.
Further, the sound source information extraction unit 10 extracts sound source information from the audio signal converted into the digital signal. The compression processing means includes a silence detection unit 16 and a silence compression availability determination unit 1 as indicated by a dashed line.
7, including the silence compression unit 18 and the like, determines whether the frame is a voice frame or a voiceless frame, and in the case of a voiceless frame, detects the number of continuous frames, and the number of continuous voiceless frames exceeds the number of reference frames for the voiced section. In this case, it is determined that silence compression is possible, and the data of the silence frames of the other consecutive numbers are converted into information on the number of consecutive silence frames, except for the number of reference frames in the consecutive number of silence frames. As the number of continuous silence frames increases,
That is, the longer the silent section is, the more the audio information can be compressed.

Further, in the output editing unit 15, the data of the sound frame including the spectrum envelope information and the sound source information, the silence data of the silent frame including the spectrum envelope information and the sound source information for the number of reference frames, and the Edit the silence frame continuation number information for silence frames excluding
The data is transferred to a voice code storage device or the like. Further, at the time of decoding, data of silence frames corresponding to the number of reference frames for a sound section remains, so that click noise can be suppressed.

〔Example〕

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 2 is a block diagram of an embodiment of the present invention, in which the same reference numerals as in FIG. 5 denote the same parts, 3 is an LSP analysis coding apparatus, 16 is a silence detecting section, 17 is a silence compression availability judging section, 18 Is a silence compression unit, which is connected to the common bus 5 and
Is controlled by The processor 6 performs various data processing.

The silence detection unit 16 includes the spectrum envelope information (LSP parameter) extracted by the spectrum envelope information extraction unit 8,
LSP composed of sound source information extracted by the sound source extraction unit 10
Based on the data, it is detected whether or not it is a silent section and the continuous number of the silent space. For example, based on the LSP data shown in FIG. 6, those having amplitude A of zero or a value close to zero are regarded as silent sections, and the number of continuous sections is counted.

Also, when sound is reproduced using the encoded sound information, if there is no sound, the previous frame (analysis section) is not referred to, but if there is sound, the LSP parameter and sound source information of the previous frame are referred to. Is what you do. That is, a sudden change in various parameters usually causes click noise, and in order to avoid this, even if silence frames are consecutive, one to several silence frames immediately before a sound section are referred to as reference frames. It would be desirable to leave it as

Therefore, the silence compression availability determination unit 17 determines that a silence section when the continuous number of silence frames exceeds the reference frame number is compressible, and when the continuous number of silence frames does not exceed the reference frame number, the silence section Is determined to be incompressible.

The silence compression unit 18 also converts the LSP data into the silence information and the continuation number information, leaving the number of reference frames immediately before the voiced section in the continuous silence frames when the silence compression availability determination unit 17 determines that silence compression is possible. And performs compression processing of the silent section. Therefore, silent frames having the number of reference frames immediately before the voiced section are output as LSP data without being compressed, and click noise is prevented as a reference frame for the voiced section during decoding. . Consecutive silence frames exceeding the number of reference frames are compressed and output. When the silence compression possibility determination unit 17 determines that silence compression is impossible due to the small number of continuous silence frames, the silence compression unit 18 does not operate.
The P data is output from the output editing unit 15 to the output terminal 2 as it is.

FIG. 3 is an explanatory diagram of silence compressed data, in which SB indicates a stop bit, T indicates a pitch period, and T1 and T0 generally indicate a frame period. For example, as described above, 2-bit "00" Although three types of frame periods are indicated by "to" 10 ", in the case of silent compression, a display that can be distinguished from the frame period, for example," 11 "is set. A is an amplitude, N is a portion corresponding to the LSP parameters ω1 and ω2 in FIG. 6 and indicates the number of continuous silences, and a portion corresponding to the LSP parameters ω3 to ω8 in FIG. 6 is all “0”. Things. In this case, the amplitude A becomes zero.

FIG. 4 shows a control flow chart of the embodiment of the present invention, and FIG.
One block (for example, 6 bytes of encoded information) is extracted from each piece of information based on the encoded information in the figure, and it is determined whether there is sound or no sound. On the other hand, if there is a sound, it is determined whether or not the previously extracted floc is silent. If there is no sound again, the counting counter is cleared and the next 1
Extract blocks.

When the sound changes from silence to sound, it is determined whether compression is possible. That is, it is determined whether or not a predetermined value, for example, three or more blocks are continuous, based on the count value of the continuous blocks with no sound.

An example of an encoded input audio block 100 and a compressed audio block 101 is shown. The encoded 6-byte, 1-block audio information a, b, c,... K,. Assuming that the block is a silence block, the silence blocks d, c,... I, j are counted in the counting step, and when changing from a silence block to a speech block, it is determined whether or not there are three or more blocks. As described above, the unvoiced blocks d, c,..., I, j are compressed into blocks i, j as shown by the compressed audio block 101 in the compression processing step.

FIG. 6 shows a case where a sound is reproduced from data obtained by such a sound encoding method and a sound is reproduced.
LSP data will be used, and if there is no sound,
It is sufficient to insert silence by the number of times N of silence continuation of the third byte of the silence compressed data shown in the figure.

Although the above-described embodiment describes the LSP system, the present invention is not limited to this system, and can be applied to other systems.

〔The invention's effect〕

As described above, the present invention includes the spectrum envelope information extraction unit 8, the sound source information extraction unit 10, and the compression processing unit, and includes a silence detection unit 16, a silence compression availability determination unit 17, and a silence compression unit 18.
By the compression means including, for example, a silent frame or a voiced frame is identified, and in the case of a voiceless frame, the number of consecutive frames is detected, and when the number of continuous frames exceeds the number of reference frames for a voiced section at the time of decoding. Is determined to be capable of silence compression, the data of other consecutive silence frames, leaving the number of reference frames in the number of consecutive silence frames, to convert the number of consecutive silence frame information, for example, As shown in FIG. 3, one frame information makes it possible to represent data of a plurality of continuous silence frames with information on the number of consecutive silence frames. There is an advantage that the storage capacity of the device can be reduced.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of the principle of the present invention, FIG. 2 is a block diagram of an embodiment of the present invention, FIG. 3 is an explanatory diagram of silence compressed data of the embodiment of the present invention, FIG. FIG. 5 is a control flow chart of the present invention in the embodiment, FIG. 5 is a block diagram of a conventional example, and FIG. 6 is an explanatory diagram of LSP data. 1 is an input terminal, 2 is an output terminal, 3 is an LSP analysis coding device, 4 is an A / D converter, 5 is a common bus, 6 is a processor (CPU), 7 is a memory (MEM), and 8 is spectrum envelope information. Extraction units, 9 and 14 are storage units, 10 is a sound source information extraction unit, 11 is an amplitude extraction unit, 12 is a pitch period extraction unit, 13 is a sound / non-speech determination unit, 15 is an output editing unit, and 16 is a silence detection unit , 17 is a silence compression availability determination unit, and 18 is a silence compression unit.

Claims (1)

(57) [Claims] A spectrum envelope information extraction unit for extracting spectrum envelope information indicating a frequency characteristic inherent to the audio signal from the audio signal converted to a digital signal; A sound source information extraction unit (10) for extracting sound source information such as an amplitude, a pitch cycle, and distinction between sound / non-speech of a sound signal; and detecting a continuous number of silence frames from the spectrum envelope information and the sound source information. When the number of detected consecutive silence frames exceeds the number of reference frames for smooth transition from silence immediately before the speech interval at the time of decoding to speech, it is determined that silence can be compressed and silence compression processing is performed. Processing means, wherein, when it is determined that silence compression is possible, the compression processing means keeps the number of reference frames immediately before the sound interval in the continuous number of silence frames as silence frames. And converts the data of the other successive silence frames except the few minutes the reference frame number of consecutive information silent frame, speech coding, characterized in that compressing the silent sections included in the audio information.