Research on Voice Interaction Robot System Based on the Hybrid Chat Model of Retrieval and Generation

With the rapid development of artificial intelligence technology, the impact of intelligent interactive robots on human life is growing. This paper constructs an intelligent voice interaction robot system based on the cloud platform and Android platform. This system uses a microphone array for audio data collection, environmental noise reduction, and voice activation. The audio processing technology based on Voice Activity Detection (VAD) and time window is adopted, which transmits data to the voice recognition engine in real time and dynamically. By invoking iFlytek’s open cloud platform’s automatic speech recognition (ASR) and synthesis services, it completes the conversion of user speech to text and the conversion of robot reply text to speech (TTS); By combining the retrieval chat model based on text similarity method and the generative chat model based on Chinese pre-training model, a retrieval-generation hybrid chat model is proposed, realizing the text communication of the robot in specific fields and daily chat needs. This voice interaction robot system can respond to users in real time, cover a wide audience, and has broad application prospects.