SPEECH PROCESSING SYSTEM AND METHOD, AND SPEECH FRAME ESTIMATING METHOD FOR REDUCING COMPLEXITY AND IMPROVING SPEECH QUALITY

摘要

PURPOSE: A speech processing system and method, and a speech frame estimating method are provided to reduce complexity and improve speech quality by optimizing a frame size. CONSTITUTION: A speech processing system includes a frame processing unit(100), a short-term analyzer(200), a short-term redundancy removing unit(250), a long-term analyzer(300), a long-term redundancy removing unit(350), and an excitation pulse search unit(500). The frame processing unit segments an input speech signal into frames and sub-frames. The short-term analyzer calculates a frame representing short-term characteristic of the input speech signal. The short-term redundancy removing unit removes the frame representing the short-term characteristic of the input speech signal and generates a noise shaped speech signal. The long-term analyzer calculates the noise shaped speech signal representing long-term characteristic. The long-term redundancy removing unit removes the noise shaped speech signal representing the long-term characteristic or frames representing the short-term and long-term characteristics of the input speech signal and generates a target vector. The excitation pulse search unit generates a pulse sequence for simulating the target vector.