Artificial Intelligence and Machine Learning Application in NPP MCR Speech Monitoring System

摘要

To explore the diversified human-machine interaction methods in the main control room (MCR) of nuclear power plants (NPP), this project established a complete set of intelligent speech control system based on Mandarin, and created a speech training system based on MFCC (Mel-Frequency Cepstral Coefficients) features and CTC (Connectionist Temporal Classification) related algorithm. This project has further enhanced the multilingual pronunciation sentences recognition rate, strengthened the logical connection between English letters, Chinese characters and numbers. Using the speech training set after training and noise reduction, when recognizing mixed sentences of Chinese characters, English letters, and numbers, this system can achieve a speech recognition accuracy rate higher than 93% and reduce the time spent on tasks by an average of 70%.