| 摘要 |
An input signal (i(t)) having an input frequency-bandwidth is decomposed into a first plurality of signals each representing a distinct, frequency-bandwidth limited portion of the input signal, and for each of the first plurality of signals a signal representing an envelope thereof is provided. The method further comprises, for each signal representing an envelope, providing a second plurality of signals (Oij(t)) each representing a distinct frequency bandwidth-limited portion of the signal representing an envelope. The envelope is preferably provided through rectification and low-pass filtering. The method is useful for analysing a human speech signal for identifying vowels in the speech signal, and for analysing a signal generated by an industrial product. <IMAGE> |
