CTC Blank Triggered Dynamic Layer-Skipping for Efficient CTC-based Speech Recognition

no code implementations • 4 Jan 2024 • JunFeng Hou, Peiyao Wang, Jincheng Zhang, Meng Yang, Minwei Feng, Jingcheng Yin

Deploying end-to-end speech recognition models with limited computing resources remains challenging, despite their impressive performance.

Knowledge Distillation speech-recognition +1

HMM-Free Encoder Pre-Training for Streaming RNN Transducer

no code implementations • 2 Apr 2021 • Lu Huang, Jingyu Sun, Yufeng Tang, JunFeng Hou, Jinkun Chen, Jun Zhang, Zejun Ma

This work describes an encoder pre-training procedure using frame-wise label to improve the training of streaming recurrent neural network transducer (RNN-T) model.

Speech Recognition

Attentive Fusion Enhanced Audio-Visual Encoding for Transformer Based Robust Speech Recognition

no code implementations • 6 Aug 2020 • Liangfa Wei, Jie Zhang, JunFeng Hou, Li-Rong Dai

The proposed method can sufficiently combine the two streams and weaken the over-reliance on the audio modality.

Robust Speech Recognition speech-recognition

The Fixed-Size Ordinally-Forgetting Encoding Method for Neural Network Language Models

no code implementations • IJCNLP 2015 • ShiLiang Zhang, Hui Jiang, MingBin Xu, JunFeng Hou, Li-Rong Dai

Information Retrieval Language Modelling +2

A Fixed-Size Encoding Method for Variable-Length Sequences with its Application to Neural Network Language Models

1 code implementation • 6 May 2015 • Shiliang Zhang, Hui Jiang, MingBin Xu, JunFeng Hou, Li-Rong Dai

In this paper, we propose the new fixed-size ordinally-forgetting encoding (FOFE) method, which can almost uniquely encode any variable-length sequence of words into a fixed-size representation.