Large-Scale Mixed-Bandwidth Deep Neural Network Acoustic Modeling for Automatic Speech Recognition
This work addresses the practical deployment issue for ASR systems by enabling unified models for mixed-bandwidth speech, but it is incremental as it builds on existing methods.
The paper tackled the problem of mixed-bandwidth acoustic modeling for automatic speech recognition by investigating strategies like downsampling, upsampling, and bandwidth extension using 1,150 hours of wideband and 2,300 hours of narrowband data, achieving performance evaluated on 8 diverse test sets.
In automatic speech recognition (ASR), wideband (WB) and narrowband (NB) speech signals with different sampling rates typically use separate acoustic models. Therefore mixed-bandwidth (MB) acoustic modeling has important practical values for ASR system deployment. In this paper, we extensively investigate large-scale MB deep neural network acoustic modeling for ASR using 1,150 hours of WB data and 2,300 hours of NB data. We study various MB strategies including downsampling, upsampling and bandwidth extension for MB acoustic modeling and evaluate their performance on 8 diverse WB and NB test sets from various application domains. To deal with the large amounts of training data, distributed training is carried out on multiple GPUs using synchronous data parallelism.