Design Of Handicap Forearm Robot Control System Based On Specific Speech Recognition

This thesis designs a control system of help-disabled forearm manipulator based on speech recognition of specific person.After training the template of speech instruction,the user can control the help-disabled forearm manipulator to realize the functions of grabbing,turning and self-balancing by voice to assist users to complete the activities required for life.Aiming at the unique characteristics of the objects used by help-disabled forearm manipulator,a speech recognition module based on isolated words is autonomously designed to realize the real-time and accurate control of the help-disabled forearm manipulator.The speech recognition module regards STM32F405 as the core chip and realizes speech recognition to a specific person through a series of processing such as acquisition,filtering,windowing,frame division,endpoint detection,feature extraction,template training and feature matching on the voice signal.Among them,the endpoint detection(VAD)based on short-term average energy and short-term average zero-crossing rate of the double-threshold method;Feature extraction using 12-dimensional Mel frequency cepstrum coefficients(MFCC parameters)and its first-order difference parameters as a feature parameter;Feature matching using improved DTW algorithm.For the chip memory used smaller,less computing power,focusing on the algorithm has been optimized.In order to meet the user to complete the end of the cup,drink and other activities to achieve hand-assisted forearm manipulator self-balancing function.The MPU6050 is used as the attitude measuring unit in the system.The quaternion-based gradient descent method is used to filter and fuse the sensor data to obtain the real-time attitude angle of the forearm manipulator.During the movement of gripping and moving the water bottle,the system controls the manipulator's hand Department always maintain level with the level of water to complete self-balancing cups and other movements.In addition,the help-disabled forearm manipulator is equipped with a pressure sensor on the linger and a machine vision module OpenMV on the wrist.When the hand grasps the target,OpenMV first identifies the target,the main control board predicts the expected pressure of the target according to the recognition result,and then adjusts the grasping force by the pressure sensor measurement data of the finger,soft grabbing of fragile objects.