Speech recognition method and client

Abstract

The invention provides a speech recognition method. The speech recognition method comprises the following steps that a client judges whether current speech recognition application involves privacy, if the current speech recognition application involves privacy, the client conducts speech recognition on collected speed signals; otherwise, the client further judges whether connection between the client and a server is successfully established; if connection between the client and the server is successfully established, the server conducts the speech recognition on the collected speech signals; if the connection between the client and the server is not successfully established, the client conducts the collected speech recognition on the collected speech signals. The invention further discloses the client of the speech recognition. The speech recognition method and the client can achieve the objective that speech recognition is conducted on speech data which contain privacy and is not limited by a network condition.

Description

technical field [0001] The invention relates to the technical field of voice recognition, in particular to a voice recognition method and a client. Background technique [0002] figure 1 Workflow diagram for existing speech recognition methods, such as figure 1 As shown, the process includes: [0003] Step 101: After establishing a connection with the client, the server receives a voice signal sent by the client. [0004] Step 102: After receiving the voice signal, the server extracts the voice recognition features in the voice signal, and performs voice recognition on the voice recognition features according to the voice recognition model. [0005] Step 103: the server returns the recognition result obtained after the speech recognition to the client. [0006] So far, the entire workflow of the existing speech recognition method has been completed. [0007] From the above analysis, it can be seen that, first of all, since the existing speech recognition is performed on...

AI Technical Summary

Problems solved by technology

[0007] From the above analysis, it can be seen that, first of all, since the existing speech recognition is performed on the server side, when the user is unwilling to upload the speech data containing privacy to the server, the speech recognition of such speech data cannot be completed; secondly, the existing The speech recognition method of the present invention is carried out on the server end, makes speech recognition be subjected to the restriction of network condition, when network condition does not support, also just can't complete speech recognition; Finally, existing speech recognition method is not adaptive to user and environment, Therefore, the server cannot provide reliable speech recognition services when there are diversity of users and the environment of users is variable.

Images