Acoustic model modeling method of residual error long short-term memory recurrent neural network

Abstract

Provided is an acoustic model modeling method of a residual error long short-term memory recurrent neural network. The method comprises: directly connecting input of a standard long short-term memory neural network framework with an internal node, splicing an original vector of the node with an input vector, and then projecting to an original dimensionality. The method provides a position of an internal node for direct connection. Experiments show that the position can bring about performance improvement of identification, and further improve system performance combined with a Dropout technology based on frames. The method improves performance of a speech recognition system based on a long short-term memory recurrent neural network, and prevents an overfitting problem combined with the Dropout technology based on frames, and realizes an objective of improving model generalization ability. The method can be widely applied in the fields of various man-machine interaction related to speech recognition.

Description

technical field [0001] The invention belongs to the field of audio technology, in particular to an acoustic model modeling method of a residual long-short-term memory cyclic neural network. Background technique [0002] With the development of artificial intelligence and computer technology, especially the development of computing hardware such as graphics processors, artificial neural networks (Artificial Neural Network, ANN) are widely used in automatic speech recognition systems. The error rate of speech recognition has also been significantly reduced with the introduction of neural networks and the increase of data sets, so it has become a research hotspot in academia and industry. [0003] The acoustic model plays an important role in the current mainstream speech recognition system, and the improvement of its performance is of great significance for improving the performance of speech recognition. Before the neural network was widely used, the basic architecture of th...

AI Technical Summary

Problems solved by technology

[0005] However, there are two major problems in the simple feedforward fully connected network: one is that the feedforward fully connected network is difficult to model the historical information of the speech signal; the other is trained in the stochastic gradient descent (Stochastic Gradient Descent, SGD) There may be problems with gradient disappearance (Vanishing Gradient) or gradient explosion (Exploding Gradient) in the process

[0008] However, in practical applications, such methods are still far from the requirements of large-scale commercialization, because the complexity of the acoustic environment still seriously affects the recognition accuracy of the speech recognition system, especially the noise resistance and robustness of the acoustic model. There is still some room for improvement

Images