Application Research Of Speech Evaluation In Embedded System Based On Characteristic Comparison

Because the calculator soft hardware technique, semi-conductor technique, electronics technique, communication technique and network technique etc. flies to soon develop, the mankind have already get into PC ages behind. The embedded equipments are usually designed for particular application, only need to identify several orders of phrasesand belong to small vocabulary quantity speech understanding system.Therefore the requirement of speech recognition don't lie in big vocabulary quantity and continuous speech understanding, , but lie in the accuracy and steady for speech recognition.This text put forward an arrangement for accomplishing speech evaluation based on Embedded System of ARM. It based on kernel chip architecture of 32 bit low power consumption and agile type. This CPU belongs to currently the ARM7 pit inside the TDMIs adopt by mostly embedded system. Micro control device of ARM has low power consumption,Low price and high performance.It is adaptable to our speech evaluation System in our text.Speech evaluation is a kind of method which is used to evaluate the difference between the following speech and the original one in the speech training. It tries to simulate the human's subjective feeling to evaluate the pitch of the input speech in order to make it consistent with the human evaluation.Speech evaluation usually use the method of phoneme evaluation.It is based on evaluation mode of speech model. It cut every phoneme unit by speech reconnition technology,and then calculate the comparison degree by comparing every phoneme unit with the phoneme Library template.The evaluation result of this method only reflect the correct degree of speech sound content,but ignore the rhythm of the process of reading text. Not only need to carry on train to a great deal of sample data with establishment speech model, also need to be ask for help of stronger operation the terrace(like PC machine) of ability carry out speech understanding function, operate quantity more and greatly, isn't suitable for embedded system.Speech evaluation method in our text is an evaluation method according to speech characteristic comparison. It evaluate the quality of a speech from a more subjective angle by comparing reference speech with standard speech. The MFCC parameter and the tone parameter have great importance to speech evaluation.So the speech evaluation method in our text discussing put the important point of evaluation on comparison of the change of rhythm. By doing this ,we can reduce the number of evaluation parameter,and low the operation quantity of evaluation operation in promising the foundation of quantity quality. This can make the speech evaluation system transplant on embedded system from the PC terrace.