The time─frequency─energy representation off speech signal in real-time recognition
Sony In · Applied Acoustics · 2001
A time- frequency- energy representation of speech signal and it s algorithm are introduced to isolated-word speech recognition. It can be characterized by two aspects: (1) non-linear time normalization, based on the gradients of short-time energy in a specific number of frequency bands, retains the transient portions and ignores the steady-state portions of speech signal in frequency domain. (2) real-time implementation due to low computational load.