JPS5941599B2 - Redundancy removed PCM audio transmission method - Google Patents

Abstract

PURPOSE:To increase the compression effect through a simple constitution and thus to obtain the excellent reproduced voice by eliminating the redundant continuation part of the vocal sound to transmit only the frame information and then giving the interpolation reproduction to the continuation part at the time of demodulation. CONSTITUTION:The PCM aural signal given from input terminal 11 with sampling of 8K Hz and quantization of 8 bits applied is divided 12 into 1-frame 80 samples (10ms). Then the signal of each frame is sent to arithmetic part 13 to undergo the calculation of the average power and the zero cross number to decide the frames for the silence, voiceless sound and vocal sound each; and these frames are added to the initial sample of each frame. While the frame-divided 12 PCM signal is supplied also to switch circuit 14. And the silence frame, voiceless frame and vocal frame are distributed fo frame eliminating part 15, but rearranging part 16 and steady frame omitting part 17 respectively. In case the changing rate is under the fixed level between the adjacent frames of the frame average power at part 17, the information showing this is sent from arithmetic part 13. And the redundant continuation parts excepting the frame information are eliminated for the subsequent frames.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a transmission system that particularly removes audio redundancy and improves transmission efficiency in audio signal transmission using PCM.

As a previously proposed redundancy removal method, TASI (
TimeAssignedSpeechInterpo
ration) method.

This system was configured to detect the presence or absence of voice during a call, and to not transmit during a silent period, and to transmit other channels during that period. In other words, this method attempts to transmit another call while the other party is talking and the other party is not producing any audio, and there is no function to compress the redundancy of the audio itself. Therefore, when the transmitted voice is continuous, it has the disadvantage that the same transmission capacity as the normal PCM system is required. Another method is adaptive bandwidth PCM.

This method divides human-generated PCM audio into frames, calculates the power spectrum bandwidth of the audio within the frame, resamples the bandwidth at a sampling rate suitable for transmission, and transmits the resulting signal. This method has the disadvantage that it is necessary to calculate the power spectrum bandwidth, which is complicated. In addition, as an improved method of the TASI method, in addition to the silent state, the determination of whether the sound is voiced or unvoiced is estimated based on the average absolute value within the frame, the number of zero crossings, and the sign change of the difference between adjacent samples. There is also a PCM with a number of quantization stages, and a voiced part is transmitted with DPCM.

This required a DPCM encoding/decoding function, and the overall configuration was complicated. The purpose of this invention is to remove the redundant continuous parts of voiced sounds, transmit only frame information, and interpolate and reproduce that part during decoding. It is an object of the present invention to provide a PCM voice transmission system capable of eliminating redundancy in a region and making effective use of communication paths.

For example, a PCM audio signal subjected to 8KH2 sampling and 8-bit quantization is input from an input terminal 11 to a frame dividing section 12 .

The dividing unit 12 divides the input signal into, for example, every 80 samples (10 mS) per frame, and adds a frame synchronization signal to each frame at the time of first sampling. The signal of each frame is sent to the calculation section 13. The calculation unit 13 calculates the sum of the squares of the values of each sample code in each frame, calculates the frame average power, calculates the rate of change in the average power between adjacent frames, that is, the rate of increase/decrease, and further calculates the sum of the squares of the values of each sample code in each frame. The number of zero crossings in each frame is calculated by counting the number of changes in the sign bit of each code. If the frame average power is sufficiently small compared to the saturation level, it is determined to be a silent frame, and the frame average power is set at 20% of the saturation level.
% or less, and the number of zero crossings is 10
In other words, if the periodic component is 1 KHz or more, the frame is determined to be an unvoiced frame, and in other cases, it is determined to be a voiced frame.

This discrimination information of silent, unvoiced, and voiced sounds is added as frame information to the first sample of each frame, similar to the frame synchronization signal. On the other hand, the frame-divided PCM signal is also supplied to the switching circuit 14, where silent frames are distributed to a silent frame removal section 15, unvoiced frames to a bit recombination section 16, and voiced frames to a steady frame switching section 17. Ru.

The silent frame removing section 15 removes all other than the frame information, and the bit recombination section 16 discards the upper 4 bits of each sampling code, and recombines the bits for two samplings into one sampling.

In other words, since the unvoiced sound has a low level and the upper bits are zero, for example, the lower 4 bits are sufficient, and the lower 4 bits of the two samples are combined. In the steady frame cutting section 17, when the rate of change of the intra-frame average power between adjacent frames is less than a constant value, that is, when there is not much change, information indicating this is sent from the calculation section 13,
Subsequent frames are stripped of all but the frame information.

The outputs of the silent frame removing section 15, bit recombination section 16, and steady frame cutting section 17 are sent to the frame connector 1.
Combined with 8. A timing generator 19 generates a clock signal and a frame signal from the clock signal from the input terminal 11, and these signals are supplied to the frame dividing section 12, the calculating section 13, and other sections. As described above, redundancy has been removed and the signal Chl, which has been shortened in time, is supplied to the multiplexing section 21.

Similarly compressed signals Chl from other channels,
Ch3, . . . are also input to the multiplexing unit B, and these signals are multiplexed so as to occupy the vacant line for 8KH2 sampling and 8-bit quantization, and are transmitted to the PCM line 22. On the other hand, on the receiving side, the signal received from the PCM line 22 is demultiplexed by a demultiplexer 23 into speech signals for each channel, frame information, and timing information, as shown in FIG.

The explanation will be focused on one separated channel signal Ehl. Frame information and timing information are detected by the timing information detection section 24 from the signal from the demultiplexing section 23. The separated channel signal Chl is sent to a frame type separator 25, which separates silent, unvoiced, and voiced frames according to control information sent from the timing information detector 24. The separated unvoiced frames are sent to a bit recombination unit 26, and the voiced frames and silent frames are supplied to interpolation units 27 and 28, respectively. The bit recombination unit 26 recombines 8 bits of one sample time into 4 bits x 2 of two sample times. In the interpolation section 27, at least 1
When a frame's worth of voiced sounds is stored and frame information indicating that steady frame truncation has been performed is input, the stored voiced sound information is read out and repeated to perform interpolation. A silent section of the silent frame signal is reproduced in the interpolation section 28.

The respective outputs of the bit recombination unit 26, the interpolation unit 27, and the outputs thereof are combined by a combination unit 29. In this way, a signal with the redundancy removed is obtained from the combining section 29. Since this demodulated signal is obtained by only processing redundant parts of the original speech, the naturalness and intelligibility of the speech are not impaired.
When a schematic temporal envelope waveform of speech is input as shown in FIG. 3, only frame information is transmitted in the silent section 31, and low bits (4
bits). In the voiced sound section 33 where the level is changing, high bits (8 bits) are transmitted for one sampling. In the voiced sound section 34 where the level change is small, only frame information is transmitted. This section 34 is interpolated by repeatedly using the immediately preceding frame 35 during demodulation. In this waveform example, the redundancy in the time occupied by sections 31 and 34 is compressed, the number of bits in the time occupied by section 32 is halved, and the redundancy is increased by 1/80 of the total as frame information. TF indicates a frame period. Note that in the above, unvoiced sounds have 8 bits per sample value without bit recombination.
Even when transmitted in bits, a large compression effect can be obtained by truncating and removing sections with small level changes in silent frames and voiced frames.

As explained above, according to the PCM audio transmission system of the present invention, redundancy contained in the audio itself is removed, the compression effect is large, and the quality of the reproduced audio is excellent.

Therefore, when applied to long-distance PCM communications, voice response devices, voice recording, etc., it has the advantage of being economical and with little deterioration in quality.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an example of the transmitting side of the redundancy-removed PCM audio transmission system according to the present invention, FIG. 2 is a block diagram showing an example of the receiving side, and FIG. 3 is a diagram showing an example of the audio envelope. It is. 11: Input terminal, 12: Frame dividing section, 13: Arithmetic section, 14: Switching circuit, 15: Silent frame cutting section, 16
: Bit recombination part, 17: Steady frame truncation part, 18
: Frame connector, 19: Timing signal generator, 21
: Channel multiplexer, 22: PCM line, 23: Channel multiplexer/demultiplexer, 24: Frame information demultiplexer, 25:
Frame type separation section, 26: Bit recombination section, 27,
28: interpolation section, 29: frame synthesis section.

Claims (1)

[Claims] 1 Divide PCM encoded audio into frames, insert frame information into each frame, distinguish each frame as silent, unvoiced, or voiced, and insert the discrimination information into each frame information, Redundancy-removed PCM audio characterized by transmitting silent frames and voiced frames whose level changes are below a certain value while leaving frame information and removing others, and interpolating and reproducing the removed portions at the time of decoding. Transmission method.