Face And Voice Mixing Kecognition System For Intelligent Ship

With the rapid development of economy,the maritime trade has become gradually prosperous,which has made the maritime traffic load heavier,the waterways more crowded and has brought greater hidden dangers to traffic.Especially when the sea Conditions are bad or the voyage is far,the driver in the process of ship's request was higher,the complex driving environment poses a huge threat to.the safety of the lives and property of the crew and passengers,and the requirements for the ship's pilot are higher,so the application of intelligent ship sailor or unmanned becomes more meaningful.As artificial intelligence and sensors develop,application of biometrics such as fingerprint recognition,face recognition,speech recognition and gesture recognition become more numerous.The use of fingerprint,voiceprint and face recognition for authentication is attracting much more attention.Among these,voiceprint recognition and face recognition are highly personalized and non-contact identification methods.The face and voice mixing recognition system for intelligent ship proposed in this paper enables the crew to use face recognition and voice recognition for dual identity authentication during driving,send instructions through voice,and complete man-machine interaction quickly and accurately.In an open and dynamic environment,dual authentication of face and voice is usually required and it ensures accuracy and security.The cabin is such an environment,which requires dual recognition of face and voice to determine the match of authorship and command execution permission.Therefore,future-oriented smart ships need this new type of intelligent human-machine interaction technology.In order to quickly and intelligently complete the human-machine interaction,it is essential to recognize the identity of the instruction sender and the command content.In view of the above problems,the main research work of this article is as follows:(1)A data-driven voiceprint feature extraction method is proposed,which uses a hierarchical clustering algorithm to perform bottom-up feature aggregation on speech data based on cepstrum and extract vectors that can represent voiceprint features from the input speech signals.And through comparative experiments,it is proved that the voiceprint features extracted by the data-driven hierarchical clustering algorithm have a better recognition effect than the classic feature by Mel-frequency cepstral coefficients(MFCC).(2)In terms of face detection,the classic face detection algorithm Haar,histogram of oriented gradients(HOG)and multitasking convolution neural network(MTCNN)are compared,it concluded that the computational performance of MTCNN is much better than the classic face detection algorithm when detecting human faces,it can also meets the needs of real-time face detection in terms of detection speed.In the aspect of face feature extraction,the local binary patterns histogram(LBPH)algorithm and the neural network based face feature extraction algorithm FaceNet were compared.On the data test set CASIA-FACEV5,FaceNet showed a huge advantage over the LBPH algorithm in accuracy.(3)Identity recognition based on voiceprint and.face recognition.Combining with the speech recognition technology of iFlytek to complete the recognition of speech content,the recognition content is processed by word segmentation to complete the exact matching of the command,and the command execution permission is verified.