Study And Design Of Specific Character Speech Recognition Based On Embedded System

Speech Recognition Technology is the process that machine understand human speech with studying. Recently recognition technology based on PC can solve the large vocabulary, speaker-independent, continuous recognition problems. With the development of embedded systems, speech recognition technology based on embedded systems has became new and important aspect of development in this field.This paper proposes the method for recognizing vehicle license speech on ARM, which can help policemen in traffic administration on-the-spot. The method is practical and can provide reference for study speech recognition based on ARM.The system has accomplished the following work:1. The whole design plan is determined by comparing several methods of speech recognition.2. The system selects ARM7TDMI processor S3C44B0X from Samsung and voice CODEC chipset UDA134ITS form Philips to implement the hardware system.3. The system adopts vector quantization and low states HMM model to reduce the requirement for memory and calculation.4. The influence of noise on dividing the words of continuous speech is analyzed in detail, and an uncomplicated and effective triphone recognition model is proposed by compress the set of observation about speech information, based on HMM theory.