Audio-Visual Efficient Conformer for Robust Speech Recognition

被引:10
|
作者
Burchi, Maxime [1 ]
Timofte, Radu [1 ]
机构
[1] Univ Wurzburg, Comp Vis Lab, IFI, Wurzburg, Germany
关键词
D O I
10.1109/WACV56688.2023.00229
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
End-to-end Automatic Speech Recognition (ASR) systems based on neural networks have seen large improvements in recent years. The availability of large scale hand-labeled datasets and sufficient computing resources made it possible to train powerful deep neural networks, reaching very low Word Error Rate (WER) on academic benchmarks. However, despite impressive performance on clean audio samples, a drop of performance is often observed on noisy speech. In this work, we propose to improve the noise robustness of the recently proposed Efficient Conformer Connectionist Temporal Classification (CTC)-based architecture by processing both audio and visual modalities. We improve previous lip reading methods using an Efficient Conformer back-end on top of a ResNet-18 visual front-end and by adding intermediate CTC losses between blocks. We condition intermediate block features on early predictions using Inter CTC residual modules to relax the conditional independence assumption of CTC-based models. We also replace the Efficient Conformer grouped attention by a more efficient and simpler attention mechanism that we call patch attention. We experiment with publicly available Lip Reading Sentences 2 (LRS2) and Lip Reading Sentences 3 (LRS3) datasets. Our experiments show that using audio and visual modalities allows to better recognize speech in the presence of environmental noise and significantly accelerate training, reaching lower WER with 4 times less training steps. Our Audio-Visual Efficient Conformer (AVEC) model achieves state-of-the-art performance, reaching WER of 2.3% and 1.8% on LRS2 and LRS3 test sets. Code and pretrained models are available at https://github.com/burchim/AVEC.
引用
收藏
页码:2257 / 2266
页数:10
相关论文
共 50 条
  • [1] Audio-visual speech recognition based on joint training with audio-visual speech enhancement for robust speech recognition
    Hwang, Jung-Wook
    Park, Jeongkyun
    Park, Rae-Hong
    Park, Hyung-Min
    [J]. APPLIED ACOUSTICS, 2023, 211
  • [2] A Robust Audio-visual Speech Recognition Using Audio-visual Voice Activity Detection
    Tamura, Satoshi
    Ishikawa, Masato
    Hashiba, Takashi
    Takeuchi, Shin'ichi
    Hayamizu, Satoru
    [J]. 11TH ANNUAL CONFERENCE OF THE INTERNATIONAL SPEECH COMMUNICATION ASSOCIATION 2010 (INTERSPEECH 2010), VOLS 3 AND 4, 2010, : 2702 - +
  • [3] Audio-visual fuzzy fusion for robust speech recognition
    Malcangi, M.
    Ouazzane, K.
    Patel, P.
    [J]. 2013 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2013,
  • [4] Research on Robust Audio-Visual Speech Recognition Algorithms
    Yang, Wenfeng
    Li, Pengyi
    Yang, Wei
    Liu, Yuxing
    He, Yulong
    Petrosian, Ovanes
    Davydenko, Aleksandr
    [J]. MATHEMATICS, 2023, 11 (07)
  • [5] Robust Self-Supervised Audio-Visual Speech Recognition
    Shi, Bowen
    Hsu, Wei-Ning
    Mohamed, Abdelrahman
    [J]. INTERSPEECH 2022, 2022, : 2118 - 2122
  • [6] Robust Audio-Visual Speech Recognition Based on Hybrid Fusion
    Liu, Hong
    Li, Wenhao
    Yang, Bing
    [J]. 2020 25TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION (ICPR), 2021, : 7580 - 7586
  • [7] AUDIO-VISUAL DEEP LEARNING FOR NOISE ROBUST SPEECH RECOGNITION
    Huang, Jing
    Kingsbury, Brian
    [J]. 2013 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP), 2013, : 7596 - 7599
  • [8] Robust audio-visual speech recognition based on late integration
    Lee, Jong-Seok
    Park, Cheol Hoon
    [J]. IEEE TRANSACTIONS ON MULTIMEDIA, 2008, 10 (05) : 767 - 779
  • [9] An audio-visual speech recognition with a new mandarin audio-visual database
    Liao, Wen-Yuan
    Pao, Tsang-Long
    Chen, Yu-Te
    Chang, Tsun-Wei
    [J]. INT CONF ON CYBERNETICS AND INFORMATION TECHNOLOGIES, SYSTEMS AND APPLICATIONS/INT CONF ON COMPUTING, COMMUNICATIONS AND CONTROL TECHNOLOGIES, VOL 1, 2007, : 19 - +
  • [10] MULTIPOSE AUDIO-VISUAL SPEECH RECOGNITION
    Estellers, Virginia
    Thiran, Jean-Philippe
    [J]. 19TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO-2011), 2011, : 1065 - 1069