Multiresolution and Multimodal Speech Recognition with Transformers
This work addresses speech recognition accuracy and efficiency for multimodal applications, but it is incremental as it builds on existing Transformer and AV-ASR methods.
The paper tackles audio-visual automatic speech recognition by using a Transformer-based architecture with multiresolution training and visual feature fusion, resulting in a 50% faster convergence, up to 18% relative improvement in word error rate over subword models, and up to 3.76% relative gain from visual information.
This paper presents an audio visual automatic speech recognition (AV-ASR) system using a Transformer-based architecture. We particularly focus on the scene context provided by the visual information, to ground the ASR. We extract representations for audio features in the encoder layers of the transformer and fuse video features using an additional crossmodal multihead attention layer. Additionally, we incorporate a multitask training criterion for multiresolution ASR, where we train the model to generate both character and subword level transcriptions. Experimental results on the How2 dataset, indicate that multiresolution training can speed up convergence by around 50% and relatively improves word error rate (WER) performance by upto 18% over subword prediction models. Further, incorporating visual information improves performance with relative gains upto 3.76% over audio only models. Our results are comparable to state-of-the-art Listen, Attend and Spell-based architectures.