Very Deep Convolutional Neural Networks For Robust Speech Recognition

Abstract

This paper describes the extension and optimization of our previous work on very deep convolutional neural networks (CNNs) for effective recognition of noisy speech in the Aurora 4 task. The appropriate number of convolutional layers, the sizes of the filters, pooling operations and input feature maps are all modified: the filter and pooling sizes are reduced and dimensions of input feature maps are extended to allow adding more convolutional layers. Furthermore appropriate input padding and input feature map selection strategies are developed. In addition, an adaptation framework using joint training of very deep CNN with auxiliary features i-vector and fMLLR features is developed. These modifications give substantial word error rate reductions over the standard CNN used as baseline. Finally the very deep CNN is combined with an LSTM-RNN acoustic model and it is shown that state-level weighted log likelihood score combination in a joint acoustic model decoding scheme is very effective

Stats

• citations44
• S2 citations—
• github stars0
• HF likes0
• heat score12.40
• arxiv keyqian2016very

Related papers

• Advances In Very Deep Convolutional Neural Networks For LVCSR (2016)0.00
• Multi-scale Octave Convolutions For Robust Speech Recognition (2019)7.16
• Analyzing Deep Cnn-based Utterance Embeddings For Acoustic Model Adaptation (2018)6.77
• Transfer Learning-based Deep Residual Learning For Speech Recognition In Clean And Noisy Environments (2025)3.58
• Residual Convolutional CTC Networks For Automatic Speech Recognition (2017)0.00
• Analyzing Large Receptive Field Convolutional Networks For Distant Speech Recognition (2019)5.84
• A Network Of Deep Neural Networks For Distant Speech Recognition (2017)10.35
• Training Augmentation With Adversarial Examples For Robust Speech Recognition (2018)11.39