Time-domain structural analysis of speech

This paper deals with an auxiliary speech signal parametrisation method based on structural analysis of speech signal in time domain. The method called TIDOSA (TIme-DOmain Structural Analysis) grounds in analysing the "shape" of incoming waveform peaks. The whole input acoustic signal is transformed into a sequence of peak shape class indices. The presented paper summarises the features of TIDOSA-based processing of speech signal, gives an overview of the executory method and the results reached so far (as the research and development is still not finished), and proposes its applications in automatic speech recognition (ASR) field.