A Speech Recognition Algorithm of Speaker-Independent Chinese Isolated Words Based on RNN-LSTM and Attention Mechanism

The speech recognition technology of isolated words is one of the widely used speech recognition technologies at present. The isolated Words Speech Recognition technology for speaker-independent is selected as the research object of the paper. And the corresponding speech database of Chinese isolated words is established. Through the research and experiment of existing speech recognition algorithms for Chinese isolated words, a model of Chinese isolated words speech recognition algorithm which based on recurrent neural networks (RNN), long short-term memory (LSTM) and attention mechanism is proposed. Attention mechanism is introduced by the algorithm model, which is used for weight adjustment, training and optimizing parameter for the combined model. The experimental results show that, compared with other algorithm models, the proposed algorithm model which based on RNN-LSTM and attention mechanism has better recognition performance for Chinese isolated words, and can effectively improve the speech recognition efficiency of Chinese isolated words for speaker-independent.