Speech Recognition Algorithm And Its Application In Embedded Applications

The research about Speech Recognition (SR) is derived since 50s, last century. After a development more than 50 years, SR techniques has been much more mature, and the scope of which has been more wider also, ranging from dictation software to language learning software on the PCs. And nowadays, the scope of SR has extended into a new arena– the embedded field.A embedded system is a custom hardware and software system, which generally refers to any dedicated computer system, which range from the various common digital devices, like cell phones, mp3 players, intelligent toys to lift, intelligent washing-machine, intelligent refrigerator etc.. They are all embedded devices. These devices have played a important role in every aspects of the human being life. As a branch of computer system, embedded systems have more effects on human being's daily life.Nowadays, the trend in the development of embedded devices is to be more convenient, more intelligent. And speech recognition techniques have played a significant role in this process. However, as to different hardware and application, different embedded devices need different software. In this paper, we discuss several common characteristic extraction algorithms, like Mel Frequency Cepstral coefficients(MFCC), Linear Predictive Cepstral Coefficient(LPCC), and several training and matching algorithms like Dynamic Time warping(DTW) and Hidden Markov Model(HMM)etc. after this, we design the speech-recognition model of an ARM9-based intelligent toy car.