BLSTM supported GEV beamformer front-end for the 3RD CHiME challenge

An analysis of environment, microphone and data simulation mismatches in robust speech recognition

“…In the following, we do not discuss DNN post-filters, which provided a limited improvement or degradation on both real and simulated data (Hori et al, 2015;Sivasankaran et al, 2015), and we focus on multichannel DNN-based enhancement instead. Table 5 illustrates the performance of the DNN-based time-invariant generalized eigenvalue (GEV) beamformer proposed by Heymann et al (2015). This beamformer is similar to the mask-based MVDR beamformer of Yoshioka et al (2015) mentioned in Section 3.2.1, except that the time-frequency mask from which the multichannel statistics of speech and noise are computed is estimated via a DNN instead of a clustering technique.…”

“…Table 8: WER (%) achieved on noisy data using various acoustic models trained on noisy real and simulated data (all channels) without sMBR (Yoshioka et al, 2015 It must be noted that, with the exception of Vu et al (2015), all challenge entrants trained GMM-HMM and DNN-HMM acoustic models on real and simulated data. Heymann et al (2015) found that discarding real data and training a GMM-HMM acoustic model on simulated data only increases the WER by 3% and 4% relative on real development and test data, respectively. This minor degradation is mostly due to the smaller size of the training set and it proves without doubt that acoustic models are able to leverage simulated data to learn about real data.…”

“…This minor degradation is mostly due to the smaller size of the training set and it proves without doubt that acoustic models are able to leverage simulated data to learn about real data. Actually, Heymann et al (2015) and Wang et al (2015) obtained a consistent ASR improvement on real and simulated data by generating even more simulated data, thereby increasing the variability of the training set. These additional data were generated by rescaling the noise signals by a random gain.…”

The third ‘CHiME’ speech separation and recognition challenge: Analysis and outcomes

“…Most systems show a large spread of performance across recording sessions (Figure 10). In particular, Hori et al (2015), Sivasankaran et al (2015), Tran et al (unpublished) and Heymann et al (2015), have overall WERs in the 9-12% range, but have WERs of 20-30% on at least one session. It is possible that these bad session scores are due to a lack of robustness to the microphone errors, given that channel errors have themselves been shown to be concentrated within sessions.…”

“…For example, Yoshioka et al (2015) apply a time-frequency mask when estimating the steering vector. Heymann et al (2015) employ a DNN to perform the necessary speech and noise covariance estimates. Other teams have employed a conventional delay and sum beamformer (e.g., Sivasankaran et al, 2015;Hori et al, 2015;Prudnikov et al, 2015).…”