Adaptive differential pulse code modulation and sub-band coding of speech signals
Adaptive differential pulse code modulation and sub-band coding of speech signals
Developments of Adaptive Differential Pulse Code Modulation (ADPCM) and Sub-band Coder for speech signals are presented that increase their robustness of transmission errors. The impetus for the study derives from the current activities in Europe to produce a continental digital cellular mobile radio system. In order to transmit digitised speech in the mobile radio environment, the speech codec must have special properties. Apart from being able to produce toll quality speech at a reasonable cost, operate at a bit rate below 24 kbit/s, they must also be able to withstand burst and distributed transmission errors, and have the capability to operate in noisy acoustical ambient conditions. Our attention is focused on two strong contenders that satisfy these requirements, namely adaptive differential pulse code modulation (ADPCM) and sub-band coding (SBC). An ADPCM codec is designed with step-size prediction and vector predictors. The step-size prediction increases the codec's dynamic range; while the vector predictors enhance the speech quality and increase the codec's robustness to transmission errors. The codec yields toll quality speech when the encoded bit rate is 24 kbit/s. The sub-band codec separates the speech signal into sub-bands by means of quadrature mirror filters and encodes each sub-band signal using one-word memory quantizers. A semi-adaptive bit allocation strategy for the quantizers is used dependent on whether the speech is voiced, unvoiced or in transition, or whether data signals are present. We have developed the SBC for operation over mobile radio channels. Specifically, we apply either step-size forcing or step-size leakage algorithms. The resultant codec configuration produces near toll quality speech while operating at transmission bit rates of 16 kbit/s. Its speech quality can be maintained without significant degradation for bit error rates (BER) up to 3x10-3. Entropy coding has been embedded into the sub-band codec. We classify the sub-band signals and the adaptive quantizer step-sizes to produce a number of codebooks. By this technique we have achieved a reduction in the average bit rate from 8% to 15%. Joint optimisation of the SBC and Reed-Solomon or convolution coding followed by post speech enhancement techniques is studied. We find that embedded systematic Reed-Solomon coding enables 16 kbit/s SBC speech to be conveyed over Rayleigh fading channels at 24 kbit/s to yield toll quality speech at BERs < 10-2.
University of Southampton
1987
Wong, Kam-Him James
(1987)
Adaptive differential pulse code modulation and sub-band coding of speech signals.
University of Southampton, Doctoral Thesis.
Record type:
Thesis
(Doctoral)
Abstract
Developments of Adaptive Differential Pulse Code Modulation (ADPCM) and Sub-band Coder for speech signals are presented that increase their robustness of transmission errors. The impetus for the study derives from the current activities in Europe to produce a continental digital cellular mobile radio system. In order to transmit digitised speech in the mobile radio environment, the speech codec must have special properties. Apart from being able to produce toll quality speech at a reasonable cost, operate at a bit rate below 24 kbit/s, they must also be able to withstand burst and distributed transmission errors, and have the capability to operate in noisy acoustical ambient conditions. Our attention is focused on two strong contenders that satisfy these requirements, namely adaptive differential pulse code modulation (ADPCM) and sub-band coding (SBC). An ADPCM codec is designed with step-size prediction and vector predictors. The step-size prediction increases the codec's dynamic range; while the vector predictors enhance the speech quality and increase the codec's robustness to transmission errors. The codec yields toll quality speech when the encoded bit rate is 24 kbit/s. The sub-band codec separates the speech signal into sub-bands by means of quadrature mirror filters and encodes each sub-band signal using one-word memory quantizers. A semi-adaptive bit allocation strategy for the quantizers is used dependent on whether the speech is voiced, unvoiced or in transition, or whether data signals are present. We have developed the SBC for operation over mobile radio channels. Specifically, we apply either step-size forcing or step-size leakage algorithms. The resultant codec configuration produces near toll quality speech while operating at transmission bit rates of 16 kbit/s. Its speech quality can be maintained without significant degradation for bit error rates (BER) up to 3x10-3. Entropy coding has been embedded into the sub-band codec. We classify the sub-band signals and the adaptive quantizer step-sizes to produce a number of codebooks. By this technique we have achieved a reduction in the average bit rate from 8% to 15%. Joint optimisation of the SBC and Reed-Solomon or convolution coding followed by post speech enhancement techniques is studied. We find that embedded systematic Reed-Solomon coding enables 16 kbit/s SBC speech to be conveyed over Rayleigh fading channels at 24 kbit/s to yield toll quality speech at BERs < 10-2.
This record has no associated files available for download.
More information
Published date: 1987
Identifiers
Local EPrints ID: 461763
URI: http://eprints.soton.ac.uk/id/eprint/461763
PURE UUID: ce07f7ca-4d79-413c-b95a-0a64d6fe5436
Catalogue record
Date deposited: 04 Jul 2022 18:54
Last modified: 04 Jul 2022 18:54
Export record
Contributors
Author:
Kam-Him James Wong
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics