Paper Info
Title
Optimized MFCC feature extraction on GPU
Abstract
In this paper, we update our previous research for Mel-Frequency Cepstral Coefficient (MFCC) feature extraction [1] and describe the optimizations required for improving throughput on the Graphics Processing Units (GPU). We not only demonstrate that the feature extraction process is suitable for GPUs and a substantial reduction in computation time can be obtained by performing feature extraction on these platforms, but also discus about the optimized algorithm. Using one GTX580 GPU our approach is shown to be approximately 97x faster than a sequential CPU implementation, enabling feature extraction to be performed at under 0.01% real-time. This is significantly faster than prior reported results implemented on GPUs, DSPs and FPGAs. Furthermore we demonstrate that multiple MFCC features can be generated for a set of predefined Vocal Tract Length Normalization (VTLN) alpha parameters with little degradation in throughput, along with the optimization for filter bank and reductions.
Year
DOI
Venue
2013
10.1109/ICASSP.2013.6639046
ICASSP
Keywords
Field
DocType
filter bank,vtln alpha parameters,vocal tract length normalization,speech recognition,throughput degradation,graphics processing units,cuda,throughput improvement,continuous speech recognition,cepstral analysis,channel bank filters,feature extraction,optimized mfcc feature extraction,computation time reduction,gtx580 gpu,mel-frequency cepstral coefficient,mfcc feature extraction,mathematical model,mel frequency cepstral coefficient,instruction sets
Mel-frequency cepstrum,Normalization (statistics),Pattern recognition,CUDA,Computer science,Cepstrum,Filter bank,Field-programmable gate array,Feature extraction,Artificial intelligence,Throughput
Conference
ISSN
Citations 
PageRank 
1520-6149
2
0.43
References 
Authors
4
4
Name
Order
Citations
PageRank
Haofeng Kou120.43
Weijia Shang234736.80
Ian R. Lane325933.64
Jike Chong413611.62