In this paper, we present an approach using the combination of artificial neural networks and time-frequency distributions to the problem of phoneme recognition in speech processing. For the inputs to the neural network, a two-dimensional Fourier transform is performed on the time-frequency distributions of the speech signals so that the resulting time-frequency pattern of a particular phoneme is always located in the same position regardless of any time and frequency shifts in the speech signal. The implementation of this approach using FFT and CPN is carried out and some preliminary results on the recognition of isolated phonemes are reported.
|Title of host publication||Proceedings of the 1995 IEEE International Conference on Neural Networks|
|Number of pages||4|
|Publication status||Published - 01 Dec 1995|
|Event||Proceedings of the 1995 IEEE International Conference on Neural Networks. Part 1 (of 6) - Perth, Aust|
Duration: 27 Nov 1995 → 01 Dec 1995
|Conference||Proceedings of the 1995 IEEE International Conference on Neural Networks. Part 1 (of 6)|
|Period||27/11/95 → 01/12/95|