Application Of Speech Recognition Technology In Navigation Device

With the rapid increase of vehicles' number and enhancement of highway network in our country,it becomes more and more urgent to locate one's own vehicles and know about the road state information.GPS navigation system satisfiy this request well.The GPS navigation system usually needs drivers to input the destination with their hands,and speech recognition technology changes this situation.As an effective interaction measures between man and machine,research about speech recognition technology is getting more and more mature and practical.With speech recognition technology applying to navigation systems,drivers can interact to it without their hands,which makes navigation and position more convenient and faster.A speech recognition system applying in GPS navigation device is designed in this paper.The research background and significance of this topic are expounded at the beginning of the paper.Then the speech recognition system and its development and present situation are introduced.And its application in the GPS navigation system is analyzed at the end of this section.The elementary theory of speech recognition is introduced in two parts,voice signal analyzing and recognition methods.Voice signal pretreatment,pronunciation vertex examination and feature extraction are introduced in the part of voice signal analyzing.Double threshold examination law based on the short-time energy and the short-time average zero crossing rates are used in vertex check-up.Also,Mel Frequency Cepstrum Coefficient parameters are selected as the characteristic parameters.The principle and the primary algorithm of HMM model and Vector Quantization model are introduced in the part of recognition methods.Discrete Hidden Markov Model is chosen as the basic recognition algorithm according to the resource-constrained characteristic of GPS navigation system.System designing,function implementation and capability analyzing of the speech recognition system which based on Discrete Hidden Markov Model are the principal part of this paper.The designing mentality and the flow of this paper are introduced in five parts,reading of pronunciation data,pretreatment,feature extraction,Vector Quantization designing and Discrete Hidden Markov Model realizing.Then the whole system is realized under lab circumstance.After that,a small-scale speech database is established under the practical application environment of the navigation system. Performance of speech recognition system and feasibility of speech recognition system using in GPS navigation system are tested and analyzed.At last the capacity of the code and the number of Discrete Hidden Markov Model state are determined considering the requirements of navigation system for correct rate,real-time,resource occupation and so on.Finally,the whole article is summarized and the inadequacy of the current work is pointed out.Also,the direction for future work is specified at the end of the paper.There are 22 figures,7 tables and 39 references in all in this article.