The Study On Speech Recognition Based On ARM

Speech recognition is one of key technologies coming into existence in field of speech processing in recent years. The purpose for this technology is to understand human language directly and make according reaction by using a kind of machine with sense of hearing like human being. The system with speech recognition function has great value in many fields such as network, communication, military, national defense etc. With rapid development of ASIC especially extremely large scale integrated circuitry, the functions of embedded system are more and more stronger. This make it possible for speech recognition technology implemented on embedded system.This paper studied speech recognition system based on ARM technology. It introduces the process of speech recognition realizing with ARM chip in theory analysis, system collectivity design, system hardware design and system software design. This paper also introduces HMM model and speech digital signal processing associated with speech recognition .Some algorithms are introduced concretely in this paper such as Mel reverse spectrum parameters ,initial value problem, data overflow problem etc. This paper analyses recognition algorithms and do some improvement to build a high performance and low operation recognition system of particular person words. All of these are experimented and based on this theory. The hardware part of this system mainly include power circuit, reset circuit, and clock circuit. The software part of this system mainly include boot loader, kernel, root file system etc. Finally, this paper talks about speech recognition technology based on ARM by a real speech recognition system including transplantation of audio driver, speech training, speech recognition and some other programs.According to the experiment, this system can recognize effectively speech .It can meet the request for being used in tough environment.