基于多任务方言识别的智能设备语音交互方法及系统
By combining deep learning and LSTM models with geolocation information, the problem of dialect recognition in complex acoustic environments for smart wearable devices was solved, achieving accurate dialect recognition and device control, and improving the flexibility and personalization of user interaction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA UNICOM ONLINE INFORMATION TECHNOLOGY CO LTD
- Filing Date
- 2025-07-31
- Publication Date
- 2026-07-17
AI Technical Summary
Existing smart wearable devices lack dialect recognition accuracy in complex acoustic environments, failing to effectively separate dialect features from the spectral overlap of environmental noise, leading to incorrect recognition of key commands. Furthermore, they cannot construct a closed-loop decision-making logic between dialect regions and user behavior, resulting in poor comprehension capabilities.
By combining geolocation information with a deep learning framework, a probabilistic correlation model between geolocation and dialect distribution is constructed, generating a fusion feature tensor of acoustic and spatial features. The LSTM model is used to capture long-term temporal dependencies, and adaptive interactive decisions are output through a multi-level decision tree to generate voice response data.
Achieve accurate dialect recognition and device control in complex speech environments, improve recognition accuracy, increase flexibility and personalization, reduce false triggering rate, and enhance user interaction experience.
Smart Images
Figure CN120766677B_ABST