Based On Talk Of Online Junk Model Verification Techniques

It is necessary to judge the credibility and exactness of the recognition result to reject the misrecognition or the word out of the vocabulary library. This is the job of utterance verification. In this paper, the word out of vocabulary and misrecognition are rejected to cut down the recognition faults, improve the system recognition ratio and reduce the false acceptance ratio.In this paper, a speech recognition system based on Hidden Markov Model is built. The basic conception and problems are introduced. The speech recognition system built for this paper is expatiated and the procedure or formulas of the train and recognition for Hidden Markov Model are brought forward step by step, and at last two tests are finished in systems for speaker independent and speaker dependent including the result and analysis.In this paper, the mathematical foundation and the ways of judgment in utterance verification are produced. The utterance verification is introduced in detail in statistical hypothesis testing and the ways of evaluation for utterance verification used in this paper in the last.In this paper, the weighted vector and formula of the gate are produced and the method with discriminant analysis is using knowledge of pattern recognition. The conceptions of on-line garbage model and N-best scores and ways of getting these two resources are introduced. Different hierarchical rejection methods are united by the discriminant function. Finally two tests are finished in systems for speaker independent and speaker dependent including the result and analysis.In this paper, Hexapod II robot speech control system is built. And this system is introduced simply. At last, a test and analysis are provided.