Design And Implementation Of Table Tennis Intelligent Speech Processing System

With the rapid development of China’s national sports industry,sports have become an indispensable part of people’s lives.Table tennis has been widely popular in China,and designing and providing a system that uses AI and related technologies to scientifically improve the skills of table tennis enthusiasts is a worthwhile research topic.This paper designs and implements an intelligent table tennis voice processing system based on the Android platform using voice and video as interaction methods.The system aims to help table tennis trainers acquire technical skills and knowledge of hitting techniques in real-time through voice dialogue during training sessions to improve their skills.The intelligent table tennis voice processing system designed in this paper includes three main parts.Firstly,in actual table tennis training scenarios,there are problems such as noisy environmental noise and mixed speech from multiple people.To solve the problem of identifying the target speaker’s voice in multiple speakers,this paper uses Mel-frequency cepstral coefficients as speaker voiceprint features.Secondly,to ensure the accuracy of the question and answer system’s response results in intelligent voice processing,the system adopts the Sim BERT semantic similarity model and is trained based on the table tennis domain’s corpus.At the same time,a statistically-based BM25 similarity algorithm and a BOOL retrieval method are introduced to reduce the time spent on semantic similarity calculations and improve the real-time interaction performance of the system.Finally,based on the design requirements,the system’s functions,including voiceprint recognition,voice wake-up,voice recognition,speech synthesis,and video retrieval and playback,are developed on the Android phone platform using existing cloud platforms and voice processing technologies.The system’s functional test results show that the system’s functions meet the expected design requirements.