APPLICATION  OF  A  NEURAL  NETWORK
TO  ON-LINE  ANALYSIS  OF  NON-STATIONARY  (SPEECH)
SIGNALS  REPRESENTED  BY  THEIR
WAVELET  TRANSFORM.
I. BASIC  PRINCIPLES

A. V. Merkusheva

Saint-Petersburg

  A property-modification detection method for non-stationary signals is suggested, that uses wavelet decomposition and a neural network. A wavelet transformation of speech signal is given using a wavelet-packet whose structure is determined by a perceptual model.