Method and device for reducing false triggering of driver's voice wake-up command words

Abstract

The invention discloses a method and a device for reducing the false triggering of a driver's voice wake-up command word, and relates to the field of automobile intelligence technology. In order to solve the problem of reducing the false triggering of the voice wake-up command word in the existing voice recognition system, the trigger sensitivity cannot be ensured. Invented based on the simpler problem of reducing the false trigger rate. The method of the present invention includes: performing voice activity detection on the received voice signal, filtering out the non-human voice region in the voice signal to obtain the filtered voice signal; obtaining the first word of the voice wake-up instruction word contained in the corresponding text of the filtered voice signal Posterior probability; calculate the confidence degree of the voice wake-up command word according to the second posterior probability of the voice wake-up command word and the voice wake-up command word; determine according to the first posterior probability and the confidence degree Whether the voice wake-up command word is the driver's voice wake-up command word. The present invention is suitable for application in the process of interaction between human and equipment through voice.

Description

technical field [0001] The invention relates to the technical field of automobile intelligence, and in particular, to a method and a device for reducing false triggering of a driver's voice wake-up command word. Background technique [0002] At present, in order to make it safer and more convenient for drivers to use in-vehicle devices, a voice recognition system is added to some in-vehicle devices, so that drivers can use voice wake-up command words to interact with in-vehicle devices without manual operation. The search voice wake-up command word is used to search, and the corresponding control voice wake-up command word is used to control the switch of the air conditioner or other equipment in the car in the car intelligent control device. [0003] In the process of realizing the present invention, the inventor found that the following problems exist in the prior art. During the actual interaction with the vehicle-mounted device through the voice wake-up command word, the...

AI Technical Summary

Problems solved by technology

[0003] In the process of implementing the present invention, the inventor found that the following problems exist in the prior art. During the actual interaction with the vehicle-mounted equipment through the voice wake-up command words, due to many interference factors in the environment, such as other people talking in the car Sound, non-human voices inside and outside the car, etc. Therefore, in the process of interacting through voice, there must be the problem of false triggering of voice wake-up command words, which will affect the performance of vehicle equipment and bring unnecessary trouble to the driver.

In order to reduce the false triggering of the driver's voice wake-up command word in the process of interacting with the vehicle-mounted device through the voice wake-up command word, two methods are usually used in the prior art: one is by increasing the recognition threshold in the voice recognition system , that is, to improve the accuracy of voice matching, which will reduce the trigger sensitivity of the voice recognition system; the other is to reduce the false trigger rate by introducing more false trigger data training models into the speech recognition system, but more false triggers It is difficult to collect trigger data, and the training modeling process for more false trigger data is more complicated, requiring more manpower and time for data processing, so the cost of this method is relatively high. In addition, in practical applications In this way, the effect of reducing the false trigger rate is not obvious

[0004] In summary, the existing methods for reducing the false triggering of voice wake-up command words in the speech recognition system cannot reduce the false trigger rate more easily on the basis of ensuring the trigger sensitivity

Images