Temporal decomposition for low rate wideband speech compression

An investigation into low bit rate wideband speech coding for applications such as unicast streaming is presented. Wideband spectral parameters are quantised below 1 kbit/s using temporal decomposition (TD) applied to the line spectral frequencies. Quantisation using TD performs significantly better than split vector quantisation at an equivalent bit rate. Disciplines Physical Sciences and Mathematics Publication Details This paper originally appeared as: Ritz, CH and Burnett, IS, Temporal decomposition for low rate wideband speech compression, Electronics Letters, 12 April 2001, 37(8), 542-543. Copyright IEEE 2001. This journal article is available at Research Online: http://ro.uow.edu.au/infopapers/154 lower performances than both the MDand the HD-based method for low Eds; (ii) the HD-based method exhibits significant degradation of the performances for an increase of Ed; and (iii) the MD-based method produces, regardless of the VAD performance, robust and superior performances in comparison with both the HDand SD-based methods. Note that for very low Eds, i.e. 0.0 I Ed 0.1, the performances of the MD and HD are slightly degraded compared with the case of Ed = 0.2. This is caused by less frequent adaptation of the noise frames due to the increased false alarm rate of VAD. In other words, VAD produces the low Ed at the expense of the increased false alarm rate at speech pauses. Experimental results using various noise sources, such as helicopter and HMMWV with levels of 0, 5, and 10 S N R dB, exhibit performance patterns similar to those shown in Figs. 1 and 2, despite differences in the absolute values being measured. Conclusion: The MD-based noise adaptation method has been proposed for robust estimation of noise variance. From the experiments, it has been shown that the MD-based method gives performances superior to both the SDand HD-based methods. 0 IEE 2001 Electronics Letters Online No: 20010368 DOZ: 10.1049/el:20010368 Y.D. Cho. K. Al-Naimi and A. Kondoz (Centre for Communication 6 February 2001 Systems Research (CCSR), University of’ Surrey: Guildford, Surrey GU2 7XH, United Kingdom) E-mail: [email protected]