Structured State Space Decoder for Speech Recognition and Synthesis
This work addresses speech processing tasks by proposing a more robust decoder, but it is incremental as it adapts an existing S4 model to new applications.
The authors tackled speech recognition and synthesis by applying a structured state space model (S4) as a decoder, achieving competitive word error rates of 1.88%/4.25% on LibriSpeech and character error rates of 3.80%/2.63%/2.98% on CSJ datasets, with improved robustness for long-form speech.
Automatic speech recognition (ASR) systems developed in recent years have shown promising results with self-attention models (e.g., Transformer and Conformer), which are replacing conventional recurrent neural networks. Meanwhile, a structured state space model (S4) has been recently proposed, producing promising results for various long-sequence modeling tasks, including raw speech classification. The S4 model can be trained in parallel, same as the Transformer model. In this study, we applied S4 as a decoder for ASR and text-to-speech (TTS) tasks by comparing it with the Transformer decoder. For the ASR task, our experimental results demonstrate that the proposed model achieves a competitive word error rate (WER) of 1.88%/4.25% on LibriSpeech test-clean/test-other set and a character error rate (CER) of 3.80%/2.63%/2.98% on the CSJ eval1/eval2/eval3 set. Furthermore, the proposed model is more robust than the standard Transformer model, particularly for long-form speech on both the datasets. For the TTS task, the proposed method outperforms the Transformer baseline.