Voice recognition method and system for multi-person speaking scene

Abstract

The invention provides a voice recognition method and a voice recognition system for a multi-person speaking scene. For the voice recognition method and the voice recognition system for the multi-person speaking scene, voice signals can be sequentially received and recognized in the multi-person speaking scene, and meanwhile, the recognition for voice collecting timestamps also can be carried outon each voice signal in the multi-person speaking scene, so that the defect that for the original voice recognition technology, the speaking sequence of different speaking ends can not be correctly restored is effectively overcome; and in addition, the method and the system can accurately restore the speaking sequence of the different speaking ends, so that the accuracy of a final voice recognition text is guaranteed, the time for follow-up voice recognition result sorting is saved, and the voice recognition efficiency is improved.

Description

technical field [0001] The present invention relates to the technical field of voice recognition, in particular to a voice recognition method and system in a multi-person speech scene. Background technique [0002] At present, speech recognition technology is widely used in the field of human-computer interaction. The existing speech recognition technology can accurately and quickly identify the speech object corresponding to the speech signal and the meaning of the speech signal itself, which greatly promotes the application of human-computer interaction. And development. [0003] However, the advantages of the existing speech recognition technology are limited to the scene where a single person speaks. For the scene of multiple speeches, speech recognition not only needs to identify the speech objects and speech meanings corresponding to different speech signals, but also recognizes and distinguishes between different speech objects. The sequence of speeches between the s...

AI Technical Summary

Problems solved by technology

It can be seen that the speech recognition method and system of this multi-person speech scene are different from the speech recognition technology of the prior art. In the multi-person speech scene, speech recognition processing can only be performed sequentially according to the order of the received speech signals. The speech recognition method and system of the present invention can not only sequentially receive and recognize speech signals in a multi-person speech scene, but also can recognize the speech collection time stamp for each speech signal in the multi-person speech scene, although the method and system There is still a time difference in the output of the corresponding speech recognition results, but the method and system can identify the speaking time points of different speaking ends according to the voice collection time stamps, thus effectively overcoming the inability of the original speech recognition technology to correctly restore the speech points of different speaking ends. The defect of speaking order, the method and system can accurately restore the speaking order between different speaking terminals to ensure the accuracy of the final speech recognition text; in addition, the method and system also sequentially process the stored speech signals by constructing a sequential buffer queue , which can not only effectively reduce the workload of speech recognition, but also ensure that each speech signal is accurately recognized and processed, thereby saving the time for subsequent speech recognition results and improving the efficiency of speech recognition

Images