LSTM time and frequency recurrence for automatic speech recognition
Authors:
- 1 December 2015
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 187-191
Abstract
Long short-term memory (LSTM) recurrent neural networks (RNNs) have recently shown significant performance improvements over deep feed-forward neural networks (DNNs). A key aspect of these models is the use of time recurrence, combined with a gating architecture that ameliorates the vanishing gradient problem. Inspired by human spectrogram reading, in this paper we propose an extension to LSTMs that performs the recurrence in frequency as well as in time. This model first scans the frequency bands to generate a summary of the spectral information, and then uses the output layer activations as the input to a traditional time LSTM (T-LSTM). Evaluated on a Microsoft short message dictation task, the proposed model obtained a 3.6% relative word error rate reduction over the T-LSTM.Keywords
This publication has 22 references indexed in Scilit:
- Constructing long short-term memory based deep recurrent neural networks for large vocabulary speech recognition2015
- Convolutional, Long Short-Term Memory, fully connected Deep Neural Networks2015
- An analysis of convolutional neural networks for speech recognition2015
- Convolutional Neural Networks for Speech RecognitionIEEE/ACM Transactions on Audio, Speech, and Language Processing, 2014
- Speech recognition with deep recurrent neural networks2013
- Deep convolutional neural networks for LVCSR2013
- Multi-dimensional Recurrent Neural Networks2007
- Learning to Forget: Continual Prediction with LSTMNeural Computation, 2000
- Long Short-Term MemoryNeural Computation, 1997
- Connectionist probability estimators in HMM speech recognitionIEEE Transactions on Speech and Audio Processing, 1994