Research And Implementation Of Information Input New Methods For The Blind

With the increasing process ability of modern computers, both the character and speech recognition technologies are developing fast, leading to the born of various recognition products. However, these high-tech products do not benefit one part of the human beings--the disabled cannot enjoy the fruit of technology development. So it is a important issue how the modern technology can serve the disable better. Research on speech recognition and braille character recognition technology and further making them into use are really meaningful and important.As for the special need of the blind to input information to certain devices, this paper researches some new methods regarding to the speech recognition and braille character recognition. A design of mobile phone for the blind based on these technology is proposed. System architecture and working principle are introduced, as well as the main parts and key interfaces.This paper also studies the speech recognition algorithm mainly based on the continuous hidden markov models and its application in speaker independent isolated word recognition systems. First the speech recognition related theories and methods are introduced from the signal preprocessing, windowing and framing, endpoint detecting, feature extraction, to the training and recognition steps. Then several speech models are compared to get a better one for the real embedded systems. In the design, a hamming window is adopted for windowing the signals, and a double-threshold method of both short-time energy and zero-passing rate is used for endpoint detecting. The mel frequency cepstrum coefficient is chosen for feature extraction, and the Viterbi and Baum-Welch algorithms are both used respectively for recognition and training processes. A kind of left-to-right continuous HMM is adopted as the speech model and critical problems including initial parameters, multi data series training and data underflow are studied carefully.A kind of complanation method for the braille character is proposed. First according to the attribute of the braille character a set of abstractions is designed so that characters can be input through touch screen. These abstractions are later segmented are features of different characters are obtained. Recognition method is the Euclidean distance. Thus, a system containing information input, character transforming and recognition is built. A embedded platform is also built based on the multi-information input methods. Software of speech processing, character processing and wireless communication are designed. These researches and designs provide a useful platform for the blind under the help of information processing new method and its application in mobile phone and some other electronic devices.This system can meet the accuracy of 92% for speech command recognition and 91% for the braille character recognition. Multi-services, such as dialing, sending messages, information acquisition, are all realized in this system on the behalf of the blind.