Performance Evaluation of a Gammatone Filterbank for the Embedded System

Abstract:

A structured gammatone filterbank is proposed to decompose the mixture acoustic signal for the embedded system. The performance of the gammatone filterbank with various filter channels is evaluated by signal to noise ratio (SNR), perceptual evaluation of speech quality (PESQ) and automatic speech recognition (ASR) accuracy. As a detailed analysis shown, the gammatone filterbank with 24 channels is good for most embedded applications with a low complex and high performance at the same time.

• Periodical: Applied Mechanics and Materials (Volumes 336-338)
• Pages: 1459-1462
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.336-338.1459
• Online since: July 2013
• Authors: Yi Jiang, Yuan Yuan Zu, Xiao Chen, Hong Zhou
• Keywords: Channel Number, Embedded System, Gammatone Filterbank

References

[1] Brown G, Cooke M. Computational auditory scene analysis. Comput Speech Lang Vol. 8(4) (1994), pp.297-336.

[2] Patterson R. An Efficient Auditory Filterbank Based on Gammatone Function. (1987).

[3] Deliang W and G. J. Brown, Slaney M. Computational. Auditory Scene Analysis: Principles, Algorithms, and Applications. NJ: Wiley and IEEE Press. (2006).

[4] Yi J, Hong Z. An algorithm combined with spectral subtraction and binary masking for monaural speech segregation [J]. ICSPCC 2011. (2011).

DOI: 10.1109/icspcc.2011.6061563

[5] ITU-T Recommendation P. 862. (2001).

[6] S. Young et al., The HTK Book, Cambridge, 3. 4. 1 edition. (2009).