Design Of Intelligent Wheelchair Control System Based On Voice And Vision

In the digital age,the contradiction between elderly people and digital devices has gradually become prominent.How to enable elderly people to adapt to a smart life is one of the urgent problems that need to be solved under the background of the aging society in the future.As one of the most convenient ways for elderly people to interact with digital equipment,nature language interaction can effectively improve the experience of elderly people using digital equipment.This thesis uses natural language or natural voice to control the intent or action of the intelligent wheelchair,and carry out the corresponding research,build and complete the control system software framework and prototype of the intelligent wheelchair.The main content of the thesis includes:(1)Introduce and adopt the encoder-decoder instruction translation model with the attention mechanism as the instruction translation core of the intelligent voice control system.Secondly,define the result form of the instruction translation model,establish a bridge between the user’s voice control and the intelligent wheelchair control system,and reasonably expand the data set required for training.Finally,based on the Keras platform and Python language,complete the development and training of the instruction model and the deployment of the cloud server to verify the feasibility of the remote voice control system.(2)By proposing quantitative indicators based on detection accuracy and reasoning speed,selecting YOLOv4 as the target detection model,and combining actual scenes to conduct targeted training on YOLOv4.On the embedded terminal,build a real-time target detector based on Tensor RT,Python and Numba.In order to combine with the voice control system,explore and complete the target real-time tracking system based on target detection and tracking.(3)Carry out the construction and functional verification of localization,navigation and control systems for intelligent wheelchairs.Aiming at the localization of the intelligent wheelchair,ORB-SLAM3 based on the ORB feature point method is selected as the core of the localization function of the intelligent wheelchair.Propose a ground segmentation algorithm combining point cloud and image,and design an obstacle perception system for intelligent wheelchairs.After integrating the localization information,target information and obstacle information,the navigation and control system of the intelligent wheelchair is designed and implemented to verify the feasibility and effectiveness of the vision-based localization,navigation,planning,and control algorithms proposed in this paper.(4)Build and verify the functions of the intelligent wheelchair experimental platform.Use Python,C++ and Boost library to develop,construct voice intelligent wheelchair control system and lower computer control program.Integrating voice control,target recognition and tracking,localization and navigation control modules,complete system-level experimental verification of intelligent wheelchairs and evaluate system performance.