Research And Application Of Voiceprint Recognition And Error Correction Technology

Voice authentication technology is a technology based on speech recognition and language models that can identify personal voice characteristics and link them to personal identity information.It has important application values in multiple fields,including personal identity verification,voice assistants,medical fields,and security monitoring.In this article,research and application of voice authentication and its error correction techniques are discussed.The main work of the paper includes:Firstly,an appropriate speech recognition model is designed through experimental analysis and research on speech recognition technology.In order to correct speech recognition results,the age information in the speech waveform is recognized to correct the speech recognition results.A convolutional neural network algorithm that classifies age as the main task is proposed,and compared with support vector machines and random forests,achieving good results.Secondly,based on the method of correcting speech recognition results using age detection,a method of detecting text information in speech recognition is proposed.By comparing the given text phrase with the speech recognition results,the effectiveness of speech recognition results is further improved.In response to errors in speech recognition results,research and analysis on the task of text correction are conducted,and a text correction model based on BERT is constructed.During the model construction process,Focal Loss is applied to optimize the error model,and the Top-k bundling method is used to improve the effectiveness of the correction model.Finally,the application of speech recognition and error correction technology in security authentication systems is discussed,and the relevant modules of different technologies are tested in actual situations.Experimental analysis proves the feasibility of the system and the effectiveness of each module,and the use of multiple technologies can improve the overall security of the security authentication system.