Method and system for adjustment of voice control accuracy

Abstract

The invention provides a method and a system for adjustment of voice control accuracy. The method is implemented by the following steps: selecting a specific neural network model, and initializing network weights; monitoring whether there is user voice control signal input; pretreating an input signal; sending the signal into the well-trained neural network after feature extraction so as to obtaina classified recognition result; judging whether the recognition result matches a voice control instruction pre-stored in the system; executing a corresponding control operation if the recognition result matches the voice control instruction; calculating the voice control accuracy if the recognition result doesn't match the voice control instruction; letting the system enter an accuracy regulation mode when the voice control accuracy is lower than a predetermined value or a user initiates the adjustment process; and retraining the network model until the accuracy is higher than the set value.The model can be trained in real time according to needs, the voice control accuracy is adjusted, and reliability of controlling a controlled plant is improved; and by using the improved neural network model, the voice recognition accuracy is raised.

Description

technical field [0001] The present invention relates to the technical field of voice control, in particular to a voice control method and system. Background technique [0002] With the development and maturity of voice recognition and human-computer interaction technology, more and more devices are controlled by voice. The controlled device obtains the user's voice signal through the voice collection equipment, compares and matches it with the voice data of the stored control command, performs voice recognition, and sends the command to the control unit after analysis to control the device to perform related operations. Among them, the accuracy of voice recognition is directly related to the accuracy of voice control, but the main problem at present is that the voice recognition rate is not high, user commands cannot be recognized or user commands are misrecognized, and there is a problem of system misoperation. [0003] The speech recognition system in the prior art basica...

AI Technical Summary

Problems solved by technology

For example, a simple problem usually requires hundreds of times or even thousands of times of iterative training, which takes a lot of time, and the network training takes a long time; another example is that the local minimum problem generally exists in the optimization process of neural networks; in the noise environment However, the performance of the existing algorithms is still not satisfactory, and the effects of existing algorithms for different noises are very different; the user speaks with an accent or uses a dialect, and the speaker's speech is vague or unclear, etc., which may also cause speech recognition effects become worse

Images