Speaker identification with wavelet decomposition and neural networks

The “cocktail party” effect describes the phenomena in which humans can selectively focus attention to one sound source among competing sound sources, which is an ability that is hampered for hearing impaired individuals. An off-line system has been developed in which a speaker is successfully identified in the presence of competing speakers for short utterances in which features used for identification are monaural, whose feature space represent a 90% data reduction from the original data. This system has also been applied to intraspeaker speech recognition. Wavelets are used to generate the multiresolution time-frequency features that are used to characterize the speech waveform. ALOPEX is an optimization paradigm that incorporates these features into a pattern recognition system through template matching or connectivity weight updating in a feedforward artificial neural network