నైరూప్య

Research on speech recognition system in complex environments

Li Mi, Xie Quanying

As for the technology progress and the development of manufacturing technique, smart devices are applied in more and more places. The usage of smart devices reduces the workload on people greatly. Besides, it expands people’s ability when work in extreme circumstances. This thesis will study the speech recognition in complex environment and acquire the effective voice message quickly in complex environment with plenty of interfering noise. Currently, most of the speech recognition products can produce voice recognition with little noise. And only in this way, they can acquire effective voice message. However, all these devices can not meet our requirements in normal environment. The speech recognition rate will drop sharp or lose efficacy in a car, which makes the current products have little adaptable range. This thesis raise a speech enhancement method who is combined with PUM model, which could eliminate the influence of noise on the speech recognition system effectively. This model improves significant recognition rate of the speech recognition system in noisy environment. It can make the speech recognition system maintain high speech recognition accuracy in vehicle's high noisy environment. By using the above technology, the speech recognition ability and the applicable scope was greatly improved. Besides, it also gives higher speech recognition accuracy than before in the same environment