CONF Motlicek_ICASSP2013_2013/IDIAP FEATURE AND SCORE LEVEL COMBINATION OF SUBSPACE GAUSSIANS IN LVCSR TASK Motlicek, Petr Povey, Daniel Karafiat, Martin EXTERNAL https://publications.idiap.ch/attachments/papers/2013/Motlicek_ICASSP2013_2013.pdf PUBLIC https://publications.idiap.ch/index.php/publications/showcite/Motlicek_Idiap-RR-37-2013 Related documents IEEE - The 38th International Conference on Acoustics, Speech, and Signal Processing (ICASSP) Vancouver, BC, Canada 2013 7604-7608 1520-6149 10.1109/ICASSP.2013.6639142 doi In this paper, we investigate employment of discriminatively trained acoustic features modeled by Subspace Gaussian Mixture Models (SGMMs) for Rich Transcription meeting recognition. More specifically, first, we focus on exploiting various types of complex features estimated using neural network combined with conventional cepstral features and modeled by standard HMM/GMMs and SGMMs. Then, outputs (word sequences) from individual recognizers trained using different features are also combined on a score-level using ROVER for the both acoustic modeling techniques. Experimental results indicate three important findings: (1) SGMMs consistently outperform HMM/GMMs (relative improvement on average by about 6% in terms of WER) when both techniques are exploited on single features; (2) SGMMs benefit much less from feature-level combination (1% relative improvement) as opposed to HMM/GMMs (4% relative improvement) which can eventually match the performance of SGMMs; (3) SGMMs can be significantly improved when individual systems are combined on a score-level. This suggests that the SGMM systems provide complementary recognition outputs. Overall relative improvements of the combined SGMM and HMM/GMM systems are 21% and 17% respectively compared to a standard ASR baseline. REPORT Motlicek_Idiap-RR-37-2013/IDIAP FEATURE AND SCORE LEVEL COMBINATION OF SUBSPACE GAUSSIANS IN LVCSR TASK Motlicek, Petr Povey, Daniel Karafiat, Martin Automatic Speech Recognition Discriminative features System Combination EXTERNAL https://publications.idiap.ch/attachments/reports/2013/Motlicek_Idiap-RR-37-2013.pdf PUBLIC Idiap-RR-37-2013 2013 Idiap Rue Marconi 19, Martigny, Switzerland November 2013 In this paper, we investigate employment of discriminatively trained acoustic features modeled by Subspace Gaussian Mixture Models (SGMMs) for Rich Transcription meeting recognition. More specifically, first, we focus on exploiting various types of complex features estimated using neural network combined with conventional cepstral features and modeled by standard HMM/GMMs and SGMMs. Then, outputs (word sequences) from individual recognizers trained using different features are also combined on a score-level using ROVER for the both acoustic modeling techniques. Experimental results indicate three important findings: (1) SGMMs consistently outperform HMM/GMMs (relative improvement on average by about 6% in terms of WER) when both techniques are exploited on single features; (2) SGMMs benefit much less from feature-level combination (1% relative improvement) as opposed to HMM/GMMs (4% relative improvement) which can eventually match the performance of SGMMs; (3) SGMMs can be significantly improved when individual systems are combined on a score-level. This suggests that the SGMM systems provide complementary recognition outputs. Overall relative improvements of the combined SGMM and HMM/GMM systems are 21% and 17% respectively compared to a standard ASR baseline.