A time frequency representations of speech signals based on a modeling of the auditory system: the gammagrams

In this paper we present a study of the temporal and spectral human auditory masking phenomena for speech signal analysis. For the purpose of modeling these masking phenomena, we used a gammachirp filterbank [T. Irino and M. Unoki, Nov. 1999], [T. Irino, 1999] to model the spectral masking and a temporal window to model the temporal masking. A global model combining these two models was built for a spectro-temporal representation. We performed a comparison of two types of spectro-temporal representations called gammagrams. The first one is based only on a gammachirp filterbank and the second is based on the global model. In addition, we performed some series of tests on different speech signals for establishing examples of masking effect curves.

[1]  Toshio Irino,et al.  An analysis/synthesis auditory filterbank based on an IIR implementation of the gammachirp , 1999 .

[2]  Toshio Irino,et al.  Noise suppression using a time-varying, analysis/synthesis gamma chirp filterbank , 1999, 1999 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings. ICASSP99 (Cat. No.99CH36258).

[3]  Hugo Fastl,et al.  Temporal masking effects: III. Pure tone masker , 1979 .

[4]  Hugo Fastl,et al.  Temporal masking effects: I. Broad band noise masker , 1976 .

[5]  A. Oxenham,et al.  Basilar-membrane nonlinearity and the growth of forward masking. , 1996, The Journal of the Acoustical Society of America.

[6]  T. Irino,et al.  A time-domain, level-dependent auditory filter: The gammachirp , 1997 .