Design And Implementation Of The Customized Voice Classification Platform

With the rapid development of Internet technology,technologies such as voice classification have gradually integrated into people's daily lives.Due to the continuous development and maturity of voice classification technology,the threshold for learning and using related technologies has been greatly improved,and it is very unfriendly for non-technical personnel who want to use voice classification technology.There is no unified solution for how to reduce the threshold for non-technical personnel to learn and use.The customized voice classification platform implemented in this thesis is aimed at voice classification,providing users with one-stop services from data preprocessing,feature engineering,model tuning,model evaluation,model prediction to result analysis.In this way,in the general scenario,ordinary business people can reach the level of advanced modelers with this platform.The core algorithm of voice classification is Google's GE2 E algorithm,and the self-attention mechanism is added to the voice classification algorithm to improve the accuracy of the algorithm.The implementation process of the platform is mainly to analyze the requirements of the platform for functions and performance,and then carry out the overall software architecture design,function module design and database design for the platform.Finally,according to the requirements analysis and platform design,a customized voice classification platform based on the Web application Flask development framework is implemented in Python language.The platform end-to-end solution is based on the Kubernetes cluster management system.The model training,evaluation and prediction are all done in one-click in the Docker container.The cloud API released by the model is based on Tensorflow Serving,in order to speed up the model inference.Through the design and implementation of the customized voice classification platform,the entire platform can run well,and all functions can be used smoothly.The platform reduces the threshold of voice classification technology,provides convenient and customized voice classification services for ordinary users,and expands the audience of voice classification technology,which contributes to the development and application of voice classification technology.