Design And Implementation Of User Defined Key Words Spotting System

Voice interaction is a very natural and convenient way of intelligent human-computer interaction.With the popularity of smart terminals,more and more devices already support voice interaction.Among the various voice interaction technologies,keyword spotting is very important.Currently,keywords are often set at the factory and cannot be changed by the user.Such devices cannot meet the needs of users with personalized command words and wakeup words.From the perspective of user-defined speech keywords,this paper studies the speech keyword spotting technology.The work in this paper covers two main areas.Firstly,a speech keyword spotting algorithm based on deep learning is implemented and improved.On the basis of the original algorithm,an attention model is introduced to improve the recognition effect of the neural network model.Then,based on the speech keyword spotting algorithm,a speech keyword spotting system that can support users to customize keywords is designed and implemented.The system generates a corresponding speech keyword detection package for the user through the speech data uploaded by the user and the user-defined voice keywords.Users only need to generate and download detection packages through this system,and then they can load different speech keyword detection packages at any time to identify different custom speech keywords.This paper introduces the speech keyword detection algorithm and improvement method used by the system in detail.Also Demonstrated and analyzed experimental results on public datasets.At the same time,the requirements analysis process,overall design and detailed design process of the whole system are described,and each functional module of the system is tested and verified.