An enhanced RASTA filtering of speech

We propose an Enhanced RASTA (ERASTA) technique for speech and speaker recognition. The new method consists of classical RASTA filtering in logarithmic spectrum domain following by another RASTA processing in spectrum domain. In this manner, both the channel distortion and additive noise are removed effectively. In isolated digit speaker identification and speech recognition experiment on TI46 database, we found that the ERASTA performed equal or better than JRASTA method in both tasks. The ERASTA does not need the speech SNR estimation in order to determinate the optimal value of J in JRASTA, and the information of how the speech degrades. The choice of ERASTA filters also indicates that the low frequency modulation components in degraded speech can deteriorate the performance of both recognition tasks. Besides, the speaker speaker needs less temporal modulation frequency band than the speech recognition.