The invention discloses a high-definition speech recognition system. The system is characterized by including a master control unit (1), a speech bank (2), a feature extraction module (3), a speech recognition module (4), an end point detection module (5), a preprocessing module (6) connected to the end point detection module (5), an audio coding module (7) connected to the preprocessing module (6), and a speech acquisition unit (8) connected to the audio coding module (7), wherein the speech bank (2), the feature extraction module (3), the speech recognition module (4) and the end point detection module (5) are connected to the master control unit (1), the preprocessing module (6) consists of a triode trigger circuit, a detection circuit and a self-excited oscillation circuit, the detection circuit and the self-excited oscillation circuit are connected to the triode trigger circuit, and the self-excited oscillation circuit is also connected to the detection circuit. The high-definition speech recognition system enables long-distance recognition to be realized. People do not need to be too close during a recognition process, and the system can recognize the people as long as the people emit sound in the distance. Meanwhile, the system is high in recognition precision and prevents unnecessary troubles for people due to recognition by mistake.