Title :
Adaptive speech coding with DCT and neutral net vector quantisation
Author :
Veleva, L.V. ; Kunchev, R.K.
Author_Institution :
Tech. Univ. of Sofia, Bulgaria
fDate :
4/15/1993 12:00:00 AM
Abstract :
A hybrid method for speech coding is presented based on the discrete cosine transform (DCT) and neural net vector quantisation (NNVQ). The neural networks are trained off-line using a two stage learning algorithm. During the encoding process, neural net adaptation is carried out. Simulation results show a high compression factor and SNR.
Keywords :
adaptive systems; decoding; discrete cosine transforms; learning (artificial intelligence); neural nets; speech coding; vector quantisation; DCT; discrete cosine transform; encoding; hybrid method; neural net adaptation; two stage learning algorithm; vector quantisation;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:19930471
