Speech recognition processing method and device

Abstract

The invention provides a speech recognition processing method and a speech recognition processing device. The speech recognition processing method comprises the steps of: acquiring data blocks to be recognized from speed data of a speech processing library, and marking recognition tags corresponding to the data blocks, wherein each recognition tag comprises a clause identifier allocated to a speech clause to which the corresponding data block is attributed according to an endpoint moment of speech detection, and a block identifier of the data block in the speech clause; sending a data block recognition request carrying the recognition tags to a speech recognition server, and receiving a data block recognition result carrying the recognition tags; and integrating a plurality of data blocks attributed to the same speech clause according to the recognition tags of all the data block recognition results, and integrating all the speech clauses. The speech recognition processing method and the speech recognition processing device achieve the effect of carrying out recognition processing on long-term continuous speech by utilizing the existing speech processing library, and improve the robustness of speech recognition.

Description

Technical field [0001] This application relates to the technical field of speech recognition processing, and in particular to a method and device for speech recognition processing. Background technique [0002] With the development of speech recognition technology, the application fields of continuous unlimited speech recognition are becoming wider and wider. For example: meeting records, recording backups, etc., often a meeting or interview takes half an hour or more, and there may be Quiet time lasting 1 minute or more. [0003] In the above application scenarios, it is necessary to use the continuous unlimited voice recognition function, and continue to work after the user turns on the voice recognition. However, the current voice recognition function is limited to short-lived voice recognition scenarios, such as voice search systems, smart home systems, etc. [0004] In the current speech recognition process, according to the resource limit of the endpoint detection speech proc...

AI Technical Summary

Problems solved by technology

However, the current speech recognition function is limited to short speech recognition scenarios, such as: voice search system, smart home system, etc.

[0004] In the current speech recognition process, according to the resource limitation of the endpoint detection speech processing library, only relatively short sentences (such as 60 seconds) can be recognized, and long-term continuous input speech cannot be recognized

Images