This article describes about modified band-pass filtered MFCC.Improved noise robustness is obtained by varying the time constants in each channel and adaptive gain compression of the envelope in each channel.This article also presents speech discrimination and
recognition results.
Mel-frequency cepstral coefficients (MFCCs),are not very robust for noise.Reasons include-the number of samples used to represent each frame is small compared to the original sampled waveform in most audio feature extraction processes.FFT followed by grouping into critical bands using triangular filters lead to discarding of information
that is not easily quantifiable.The temporal information in the signal is distributed in the magnitude and phase of the multiple frequency bins.Masking of pertinent information
is the result of their combination and gives a representation that has low time and frequency resolution.
Asymmetrical shape of the cochlear filters allow for good time resolution and good frequency resolution in human auditory system.Thus,Noise robust auditory features (NRAF)
has been developed to solve such problems.These are based on a model ofthe human auditory system the computational costs are less.

The report pdf can be downlaoded at:
http://mediafire?8xpvc78z5jmssaq