Robust Speech Recognition Technology Optimization In Complex Spatical Noise Environment

With the improvement of social productivity and the rapid development of science and technology,people become more and more concerned about human-computer interaction.However,in the process of actual human-computer interaction,there are many unstable factors in speech recognition due to the uncertainty of environment and background noise.The environment of noise as example,it is necessary to analyze the speaker the dialogue,as well as solve the environmental noise caused by the speaker sound masking effect and other issues.In the meanwhile,to meet the real-time and reliability requirements,multiple other factors,all remixed to effect the application of acoustic recognition in limited space environment.At present,the optimization of speaker robust speech recognition in complex space noise environment has attracted much attention of related researchers.To solve speech recognition in complex space noise environment,a robust speech recognition technique is proposed in this paper.The features of application environment scheme combined with complex space,especially for indoor environment.The application of long space theory of composite sound source determination method of noise model is established,and a combination of scene layout,the optimization compensation algorithm improves the adaptability to the characteristics of the scene;in speech recognition application scenarios,before the level of voiceprint recognition adopt multi-channel sampling method to determine the number of speakers;improved recognition system for speaker number smooth decision making ability,at the same time the new adaptive characteristics to improve its adaptability to environment.Also with noise characteristics and factors of sampling equipment layout scene characteristics,of the adaptive noise scenarios model matching and training using support vector machine method,can effectively improve the success rate of speech recognition site.The recognition system has been proved in small scale laboratory.Tested in Home Furnishing indoor environment,the embedded host platform recognized the speech and speaker verification.The implementation case has been partially applied to the project,effectively improving the interactive experience of the original end user.The research and practice of this subject,to provide an effective solution for robust speech recognition optimization technology of complicated spatial noise environment,and the future of the embedded platform based on the design robustness in different noise scene has provided valuable experience accumulated.