Reference Matlab/Octave implementations of feature extraction algorithms

The scripts provided in this software package were written to perform the feature extraction in automatic speech recogniton experiments and to evaluate the obtained recognition performance in [1].

This includes the calculation of:

• Logarithmically scaled Mel-spectrograms (LogMS) in log_mel_spectrogram.m
• Mel-frequency cepstral coefficient (MFCC) features in mfcc_feature_extraction.m
• Gabor filter bank features (GBFB) features in gbfb_feature_extraction.m
• Separable Gabor filter bank (SGBFB) features in sgbfb_feature_extraction.m
• Histogram equalization (HEQ) in heq.m
• Equal-performance SNR increase (EPSI) in epsi.m

Detailed explanations of the corresponding concepts are provided in [1]. The SGBFB feature extraction is closely related to the GBFB feature extraction which was introduced in [2]. For an overview of current publications and further developments of the SGBFB front-end, visit http://medi.uni-oldenburg.de/SGBFB

play_demo.m is demonstrates the use of the feature extraction scripts.

References

[1] M.R. Schädler and B. Kollmeier, "Separable spectro-temporal Gabor filter bank features: Reducing the complexity of robust features for automatic speech recognition", J. Acoust. Soc. Am. Volume 137, Issue 4, pp. 2047-2059, DOI: 10.1121/1.4916618, URL: http://link.aip.org/link/?JAS/137/2047 (2015)

[2] M.R. Schädler, B.T. Meyer, B. Kollmeier "Spectro-temporal modulation subspace-spanning filter bank features for robust automatic speech recognition", J. Acoust. Soc. Am. Volume 131, Issue 5, pp. 4134-4151, DOI: 10.1121/1.3699200, URL: http://link.aip.org/link/?JAS/131/4134 (2012)