GB2117608A

GB2117608A - Speech signal transmission system

Info

Publication number: GB2117608A
Application number: GB08304225A
Authority: GB
Inventors: Tad Weng Chong; Peter Bylanski
Original assignee: General Electric Co PLC
Current assignee: General Electric Co PLC
Priority date: 1982-02-17
Filing date: 1983-02-16
Publication date: 1983-10-12
Also published as: GB2117608B; GB8304225D0

Abstract

In a wide band PCM transmission system an 8 KHz bandwidth speech signal is sampled at 16 KHz and alternate samples are PCM coded and transmitted over a standard 64 Kbit/s channel. Predictor coefficients for the missing samples are derived at the transmitter end for transmission over the 64 Kbit/s channel, for example by "bitstealing". In an alternative arrangement where sampling is at 8 KHz, the receiver may have means to interpolate a sample between each pair of received samples either by simple interpolation or with the one of predictor coefficients which may be derived from the received samples or be fixed. <IMAGE>

Description

SPECIFICATION Speech-signal transmission systems The present invention relates to speech-signal transmission systems.

In particular although not exclusively the invention relatestospeech-signal encoding arrangements for pulse-code modulation transmission systems, and to transmission systems incorporating such encoding arrangements.

Present pulse-code modulation, or PCM, telephone systems utilise speech frequency bandwidths of about 3,4 KHz,the speech signal amplitude being sampled at a rate of 8KHz and the sampled amplitude values encoded using an eight-bit code. The encoded speech signal therefore takes the form of a stream of binary digit values at a rate of 64K bits/sec, and local and trunktransmission systems are commonly arranged to accommodate interlaced or multiplexed digit streams at rates which are multiples of 64K bitsisec.

With the advent of electro-acoustictransducers capable of handling speech signals of greater bandwidths than the standard 3.4 KHz improvements in the encoding are required to take full advantage ofthe better speech quality offered. One way of achieving this in a pulse-code modulation system would be to sample the speech signals at a higher rate, say 16 KHz, thus enabling a speech bandwidth of up to 8 KHz.

With eights bits per sample this would, however, require a 128K bit/sec transmission channel ortwo standard 64K bit(sec channels, whereas it would clearly be advantageous if improved quality speech could be transmitted over a single 64K bit/sec channel.

According to one aspect of the present invention in a speech-signal encoding arrangement in which the instantaneous value of an analogue speech signal is arranged to be sampled at regular intervals and respective groups ofdigitvalues derived which represent the respective values of at least some of said samples, there are provided means to derive from a plurality of said samples orfrom the respec tivegroupsofdigitvaluesa plurality of digitally encoded coefficient values from which a succeeding samplevalueorgroup ofdigitvalues may be predicted, and means to provide an output pulse train incorporating some at least of said groups of digit values and some at least of said digitally encoded coefficient values.

Preferably said output pulse train incorporates the respective groups of digitvatuesforalternatesam- ples.

The digitally encoded coefficient values may be incorporated in said output pulsetrain in addition to said groups of digit values or in place of predeter mined digit values of said groups of digit values.

According to another aspect of the present invention in a pulse-code modulation speech signal transmission system there are provided at a transmitting station means to derive from speech signal sample values digitally encoded coefficientva lues from which succeeding speech signal sample values may be predicted and means to transmit said encoded coefficient,va lues tog ethe r with some at leastof the coded sample values, and there are provided art a receiving station means responsive to received sample values and received coefficient values to derive predicted speech signal sample values for interpolation between received sample values.

According to another aspect of the present invention in a pulse code modulation speech transmission system comprising a transmitting station and a receiving station there are provided atthetransmitting station means to sample the instantaneous value of a speech signal at regular intervals, means to pulse code the values of some at least of these samples, meansto derivefrom a plurality of samplevalues a set of digitally coded coefficient values from which a value for a succeeding sample may be predicted, and means to transmit coded signals in respect of alternate samples together with said coded coefficientvalues, and there are provided atthe receiving station means responsiveto the values of said alternate samples and ta said coefficient values to derive predicted sample values for interpolation between the received sample values.

The coefficientvalues may remain substantially unchanged for a plurality of sample intervals,for example for periods often milliseconds or more. The coded coefficient values can therefore be transmitted at a much lower rate than for the coded speech signals, for example by "bit stealing" in place of, say, the least significant bits of a plurality of coded samples.

A pulse code modulation speech-signal transmission system incorporating an encoding arrangement in accordance with the present invention will now be described with reference to the accompanying draw ings, ofwhich:- Figure 1 shows in blockschematicform a speechsignal encoding arrangement for atransmitter station of the system, and Figure 2 shows in block schematic form a decoding arrangement for a receiver station of the system.

Referring first to Figure 1 at a transmitting station of a wideband speech transmission system a speech signal from a microphone transducer (not shown) are applied bywayofa low-pass filter 1, having a cutoff frequency of say 8 KHz ratherthan the usual 3.4KHz, to a coder 2, where the instantaneous value of the speech signal is sampled ata rate up to, say, 16 KHz and coded in accordance with an eight bit code such as the CCITTA-law code. The coder 2 therefore provides as an output signal a stream of binary digit values or bits at a rate of 128K bit/sec, which are held by a delay circuit 3 while predictor coefficients are derived in known manner by an anlyser4.

Coded samples leaving the delay circuit 3 are passed or blocked by a gate 5 to provide an average sample rate of 8 KHzto a side-channel insertion circuit 6, where predictor coefficientvalues are combinedwith the coded speech samples. Abuffer7 is provided so that the combined gated samples and coefficient values can be transmitted as a continuous stream of bits at 64K bits/sec.

Depending on the initial sampling rate the coded samples may be passed and blocked in groups by the gate 5 to provide the required average rate, with the predictor coefficients being intended to provide predicted values at a receiving station forthose samples that are blocked. In the particular case where the initial sampling rate is 16 KHzthe passing of alternate coded samples by the gate 5 provides the necessary average transmitted sample rate while allowing decoding by a standard PCM decoder.

Taking as an example third-order prediction of blocked samples, the values of three coefficients must be transmitted with the coded speech samples at a rate sufficient to allow for updating in accordance with changes in the speech waveform. It is known that speech spectral features may reasonably be assumed to be constant over frame intervals of, say, 10 m secs.

If the predictor coefficients each have 4-bit resolution the side channel insertion circuit 6 will receive 12 bits every 10 m secs. Allowing, say, 8 bits for framing information a total of 20 bits mustthen be transmitted every 10 m secs, for example by replacing or "stealing" the least significant bit of every fourth coded sample transmitted.

Referring to Figure 2,thetransmitted signal at 64K bit/sec is first applied to a side-channel extractor circuit8,wheretheframing pulses and coefficient values are removed from the stream of coded speech samples and passed to a predicted-sample synthesiser9. The coded samples are then held in a buffer or delay circuit 10 while the predicted sample values to be interpolated are synthesised, the interpolation then being carried out by a gate circuit 11. The transmitted and predicted samples are subsequently decoded and filtered in the normal way but at the higher sample rate.

While the predictor coefficients may not enable exact reconstruction of the waveform ofthe speech signal it should enable sufficient reconstruction ofthe frequency spectrum of the speech signal to give a subjective quality close to thatwhich would be obtained with ungated 16 KHz sampling.

In an alternative arrangement where speech signals in a frequency band extending to 8 KHz are sampled art a rate of 8 KHzfortransmission over a standard 64 Kbit/sec channel, the receiving station may be provided with means to interpolate, say, one extra sample value between each received value and the next. The values of these extra samples may be derived simply as the mean value of the received samples between which they are to be interpolated, orthe received samples may be analysed periodically to derive predictor coefficients. Alternatively a fixed set of coefficients may be utilised forthe prediction of the sample values two be interpolated.

According to another aspect of the present inven tiontherefore in a pulse-code modulation speech signal transmission system comprising a transmitting station and a receiving station there are provided at the transmitting station means to sample the instantaneous value of a speech signal at regular interva Is, means to pulse code the values of these samples, and there are provided at the receiving station means to interpolate a sample value between consecutive received sample values in dependence the values of said consecutive received samples or in dependence uponthevalues of a predetermined selection of samples.

Preferably sample values are interpolated between each received sample and the next received sample.

Claims

1. Aspeech-signal encoding arrangement in which the instantaneous value of an analogue speech signal is arranged to be sampled at regular intervals and respective groups of digit values derived which represent the respective values of at least some of said samples, wherein there are provided means to derivefrom a pluralityof said samples orfromthe respective groups of digit values a plurality of digitiallyencoded coefficient values from which a succeeding sample value or group of digit values may be predicted, and means to provide an output pulse train incorporating some at least of said groups of digit values and some at least of said digitally encoded coefficientvalues.

2. Aspeech-signal encoding arrangement in accordance with Claim 1 wherein the output pulse train incorporates the respective groups of digit valuesforalternatesamples.

3. Aspeech-signal encoding arrangement in accordance with Claim 1 or Claim 2 wherein the digitally encoded coefficient values are incorporated in said output pulse train in addition to said groups of digit values.

4. Aspeech-signal encoding arrangement in accordance with Claim 1 or Claim 2 wherein the digitally encoded coefficient values are incorporated in said output pulse train in place of predetermined digit values of said groups of digitvalues.

5. A pulse-code modulation speech signal transmission system wherein there are provided at a transmitting station means to derive from speech signal sample values digitally encoded coefficient values from which succeeding speech signal sample values may be predicted and means to transmit said encoded coefficient values together with some at least of the coded sample values, and there are provided art a receiving station means responsive to received sample values and received coefficient values to derive predicted speech signal sample values for interpolation between received sample values.

6. A pulse-code modulation speech signal transmission system comprising a transmitting station and a receiving station, wherein there are provided at the transmitting station means to samplethe instantaneous value of a speech signal at regular intervals, means to pulse code the values of some at least of thesesamples, means to derive from a plurality of samples values a set of digitally coded coefficient values from which a value for a succeeding sample may be predicted, and means to transmit coded signals in respect of alternate samples together with said coded coefficient values, and there are provided at the receiving station means responsive to the values of said alternate samples and to said coeffi cient values to derive predicted sample values for interpolation between the received sample values.

7. A pulse-code modulation speech signal transmission system in accordance with Claim 6 wherein the coefficient values remain substantially unchanged for periods of the order often milliseconds.

8. A pulse-code modulation speech signal transmission system comprising a transmitting station and a receiving station wherein there are provided at the transmitting station means to sample the instan taneousvalue of a speech signal at regular intervals, means to pulse code the values of these samples, and there are provided at the receiving station means to interpolate a sample value between consecutive received sample values in dependence the values of said consecutive received samples or in dependence upon the values of a predetermined selection of samples.

9. A pulse-code modulation speech signal transmission system in accordance with Claim 8wherein a plurality of predictor coefficients are arranged to be derived in dependence upon the values of said predetermined selection of samples, said coefficients being utilisedforthepredictionofthesamplevalues to be interpolated.

10. Apulse-code modulation speech signal transmission system in accordance with Claim 9wherein said predictor coefficients are arranged to be derived at said transmitting station.

11. A pulse-code modulation speech signal transmission system in accordance with Claim 9 wherein said predictor coefficients are arranged to be derived at said receiving station in dependence upon the values of a predetermined selection of received samples.

12. A pulse-code modulation speech signal transmission system in accordance with Claim 8wherein a sample value interpolated between consecutive received sample values is substantially the mean value of those two consecutive received sample values.

13. A pulse-code modulation speech signal transmission system in accordance with Claim 8 wherein each samplevalueto be interpolated is arranged to bederivedin dependence upon the values of a predetermined selection of received samples and a predetermined set of predictor coefficients.

14. A pulse-code modulation speech signal transmission system in accordance with Claim 8wherein one or more sample values are interpolated between each received sample and the next received sample.

15. A pulse-code modulation speech signal transmission system substantially as hereinbefore de scribedwith reference to the accompanying drawing.