It is a technology that converts speech signals into corresponding texts or commands through recognition and conversion. Speech recognition mainly includes speech input, feature extraction, speech modeling and pattern matching.
The essence of speech recognition can be summed up as the process of pattern matching and pattern recognition, which is generally divided into two steps: the first step is to establish the analyzed speech pattern and the acoustic pattern of the recognition unit; The second stage is speech recognition and testing. According to the mode of the existing system, the corresponding algorithm is selected, the required feature parameters are extracted according to the speech analysis scheme, and then compared with the system model, and the recognition result is obtained through judgment.
Speech recognition is divided into specific person speech recognition and non-specific person speech recognition. Specific person voice recognition technology is aimed at specific vocal recognition technology. Simply put, it can only recognize one person's voice, which is not suitable for a wider group. Speaker-independent speech recognition technology can meet the voice recognition requirements of different people and is suitable for a wider range of crowd applications. In recent years, the development of speaker-independent speech recognition chip makes the control of man-machine interface equipment based on speaker-independent speech recognition more simple and flexible, and the purpose of speaker-independent speech recognition can be achieved through the processing of microprocessor chip.