OCR-Enhanced Multimodal ASR Can Read While Listening

TL;DR

This paper introduces Donut-Whisper, a multimodal ASR model that leverages visual subtitles and audio to improve speech recognition in English and Chinese, demonstrating significant performance gains over baselines.

Contribution

It proposes a novel audio-visual ASR model with dual encoders, a cross-attention module for modality alignment, and a lightweight knowledge distillation scheme, along with a new multilingual dataset.

Findings

Achieved 5.75% WER reduction on English

Achieved 16.5% CER reduction on Chinese

Demonstrated superior performance over baseline models

Abstract

Visual information, such as subtitles in a movie, often helps automatic speech recognition. In this paper, we propose Donut-Whisper, an audio-visual ASR model with dual encoder to leverage visual information to improve speech recognition performance in both English and Chinese. Donut-Whisper combines the advantage of the linear and the Q-Former-based modality alignment structures via a cross-attention module, generating more powerful audio-visual features. Meanwhile, we propose a lightweight knowledge distillation scheme showcasing the potential of using audio-visual models to teach audio-only models to achieve better performance. Moreover, we propose a new multilingual audio-visual speech recognition dataset based on movie clips containing both Chinese and English partitions. As a result, Donut-Whisper achieved significantly better performance on both English and Chinese partition of the dataset compared to both Donut and Whisper large V3 baselines. In particular, an absolute 5.75% WER reduction and a 16.5% absolute CER reduction were achieved on the English and Chinese sets respectively compared to the Whisper ASR baseline.