Cabin Chinese Voice Assistant Control Command Dataset

Currently, smart driving technology is becoming increasingly widespread, but there is significant room for improvement in the accuracy and reaction speed of its voice assistants. In-car voice control systems often encounter issues of misrecognition and incorrect reactions, leading to poor user experience. Some existing solutions have very limited recognition rate and response speed due to lack of optimization for the driving environment. This dataset aims to improve voice recognition accuracy, response speed, and stability in noisy environments through a rich sample of voice commands. Data collection is conducted using high-fidelity audio equipment in the cabins of various vehicle models, covering multiple traffic environments such as urban, suburban, and highways; quality control is stringent, including multi-round manual annotation, consistency verification, system evaluation, and expert review; the annotation team consists of traffic engineering experts and voice technology engineers, totaling 30 people; data preprocessing involves background noise reduction, audio segmentation, and voice enhancement; the final data is stored in WAV format, organized by environment and command type.