Improved MVDR Beamforming Using LSTM Speech Models to Clean Spatial Clustering Masks
This work provides an incremental improvement in speech enhancement for speech recognition systems operating in noisy multi-channel environments.
This paper integrates LSTM speech models with spatial clustering to improve multi-channel noise reduction. The proposed system, when applied to the CHiME-3 dataset, increased speech quality (PESQ) and reduced the word error rate of the baseline speech recognizer compared to the default BeamformIt beamformer.
Spatial clustering techniques can achieve significant multi-channel noise reduction across relatively arbitrary microphone configurations, but have difficulty incorporating a detailed speech/noise model. In contrast, LSTM neural networks have successfully been trained to recognize speech from noise on single-channel inputs, but have difficulty taking full advantage of the information in multi-channel recordings. This paper integrates these two approaches, training LSTM speech models to clean the masks generated by the Model-based EM Source Separation and Localization (MESSL) spatial clustering method. By doing so, it attains both the spatial separation performance and generality of multi-channel spatial clustering and the signal modeling performance of multiple parallel single-channel LSTM speech enhancers. Our experiments show that when our system is applied to the CHiME-3 dataset of noisy tablet recordings, it increases speech quality as measured by the Perceptual Evaluation of Speech Quality (PESQ) algorithm and reduces the word error rate of the baseline CHiME-3 speech recognizer, as compared to the default BeamformIt beamformer.