本作品設(shè)計(jì)了一種基于嵌入式技術(shù)的智能型殘疾人輪椅，此系統(tǒng)以Intel core duo 2微處理器為核心，以Window XP為開發(fā)平臺(tái)，運(yùn)用語音識(shí)別與合成、圖像處理與特征提取、基于代數(shù)算法神經(jīng)網(wǎng)絡(luò)的模式識(shí)別等技術(shù)，實(shí)現(xiàn)了智能輪椅的語音控制與交互、障礙物檢測、任意物體識(shí)別等功能。該智能輪椅使得殘疾人可以通過語音來控制輪椅的運(yùn)動(dòng)狀態(tài)；使用激光測距，使得智能輪椅在規(guī)定距離內(nèi)有效避開障礙物；通過基于紋理的圖像分割算法對(duì)圖像進(jìn)行區(qū)域分割，實(shí)現(xiàn)對(duì)分割區(qū)域顏色特征和形狀特征的提取。再將提取的顏色特征和形狀特征作為神經(jīng)網(wǎng)絡(luò)輸入，完成對(duì)物體的識(shí)別。給嚴(yán)重視力障礙的人“看見”物體帶來一線希望。 

        其中所使用的圖像分割算法是應(yīng)用自創(chuàng)的基于紋理特征和邊緣特征的圖像分割算法，較大精度地將圖像中的物體分割出來并對(duì)每個(gè)區(qū)域進(jìn)行顏色特征和形狀特征的提取。

        運(yùn)用一門前沿的計(jì)算科學(xué)——人工神經(jīng)網(wǎng)絡(luò)來實(shí)現(xiàn)模式識(shí)別，將提取到的圖像區(qū)域的顏色特征和形狀特征當(dāng)作神經(jīng)網(wǎng)絡(luò)的輸入，從而識(shí)別物體。在算法的選擇上，摒棄了在應(yīng)用中占統(tǒng)治地位的BP神經(jīng)網(wǎng)絡(luò)算法，將近年提出的基于代數(shù)算法的神經(jīng)網(wǎng)絡(luò)理論成功的用程序?qū)崿F(xiàn)，并且證明了這個(gè)理論的可行性和優(yōu)越性，是一次大膽而且成功的嘗試。

        該智能輪椅給殘疾人用戶帶來行動(dòng)上的更大靈活性，具有較高的實(shí)用價(jià)值和廣闊的應(yīng)用前景。

An idea of intelligent wheelchair based on embedded technology is proposed. Based on the microprocessor of Intel core duo 2 and the development platform of Window XP, and using the technology of speech recognition and synthesis, image processing and characteristics extraction, and pattern recognition based on algebraic algorithm for neural network, the intelligent wheelchair completes the functions of speech control and interaction, obstacle detection, object recognition and so on. This intelligent wheelchair makes the disabled people control the move statement through the speech control technology, and completes the object recognition function to ‘see’ the objects for the disabled.

The image partition algorithm used in this system is based on the grain characteristics and edge characteristics, and is designed by ourselves. It can divide up the object and distill the color and figure characteristics of every part.

This system uses artificial neural network which is a kind of foreland numeration to realize pattern recognition. It viewed the color and shape characters of the sampled image area as input signal of artificial neural network to recognize objects. Although the BP neural network algorithm dominates the applications, we choose the recently proposed neural network based on algebraic algorithm, realized it in our program and verified the feasibility and advantage of this algorithm. It is a brave and successful experiment. 

This brings great convenience for the disabled and has promising future in applications as well as high practical value.

linux操作系統(tǒng)文章專題:linux操作系統(tǒng)詳解（linux不再難懂）