Research On Speech Control Technology Based On Semantic Comprehension

Voice Control technology is the application of speech recognition theory in the field of control,with great significance of practical research value and application significance.It requires not only a computer can correctly identify the voice commands,but also requires that it can respond to the voice instructions to achieve the purpose of the freedom of man and machine interaction.Therefore,from the perspective of speech and language,voice control technology was studied.In the level of acoustic signal,it analyzed the endpoint detection and its application problems in complex noisy background.Double threshold endpoint detection algorithm lost its accuracy in low SNR,and therefore endpoint detection algorithm combined with Empirical Mode Decomposition(EMD)and cross-entropy was proposed.Simulation experiments on the platform of Matlab indicateed that the validity of the method in complex noise environments.In the level of selection control scheme,keyword systems,which are more intelligent than isolated word systems and faster than continuous speech recognition system,were choose because of its superiority,and according to the features of natural language,design ideas of keyword systems based on LVCSR were selected.As the research background of the smart car system,keyword detection experiment conducted with its keyword list.By analyzing the lack of results of the keyword detection experiment,keywords were segmented,and keywords table were re-establish.Detected multiple keywords at the same time and finally,the result of detection were amended.After Improving,the accuracy of keyword detection was greatly increased.On the other hand,it detected the keywords while completed the segmentation of Chinese word so that it greatly reduced the workload of semantic understanding module,laying a good foundation.In the level of analysis of the command statements,the results of keyword detection were studied and discussed for further.Because the characteristics of the command statements of control system are mostly imperatives which are centered on the word of the verb,the Case Grammar were used in the analysis.Combined with the application background,the Case Grammar were analyzed with its application and were proposed for ideas as improving the definition of the "direction case",amendingthe speed case concept and merging the cases of time,length and angle for “added case”.Filled case semantic framework,and converted it into machine instructions.Finally,the results of system experiments were analyzed and discussed.