Design And Implementation Of Identity Recognition System Based On Combination Verification Of Face And Voice Pattern

At present,a lot of research results have been obtained and widely applied based on single-mode biometric technology.However,with the continuous expansion of application fields,single-mode biometric identification technology has shown many drawbacks and limitations in application.Specifically,in complex environments,certain biological features are vulnerable to interference in the acquisition process.The progress of biometric forgery technology makes the singlemode identification system have potential safety hazards,which affects the wide spread of biometric identification technology.The paper illustrates the design and implementation of an identification system based on face and voiceprint combination by studying face recognition algorithm,voiceprint recognition algorithm and multimodal biometric fusion algorithm.In the area of face recognition algorithm research,a median filter is used to remove information such as noise in the image to resolve the issue that the face image will be interfered by environmental factors during the imaging process and affect the image quality.As different regions of the face image contribute differently to the recognition rate,an improved HOG algorithm based on inverse different moment weighting is introduced.The face image is divided into blocks,and the inverse different moment of the image is used as the weight.The optimal weight coefficient of each sub-block is calculated,which effectively improves the recognition rate of HOG features.Feature dimensionality reduction is realized by adopting PCA algorithm to address the issue of HOG's high-dimensional feature vectors.While conducting research on voiceprint recognition algorithm,the dual threshold endpoint detection method is used to eliminate the silent segment so as to address the mute problem of the voice signal during the recording process.As MFCC coefficient can only describe static characteristics,a study was conducted on first order differential MFCC coefficient(MFCC)to effectively compensate for the deficiency of dynamic characteristics of MFCC.In order to solve the problem that the number of voice frames with MFCC features is not fixed,GMM-UBM and I-Vector were used to cluster MFCC features and reduce dimensionality.In order to address the problem that single-modal biometric system has low recognition accuracy and safety factor,a feature extraction method based on HOG feature and MFCC feature series connection and weighted fusion was adopted.In terms of the hardware in the system design,i.MX6 Q was used as the central processor,and Le Eco somatosensory camera and Raspberry Pi microphone as the data acquisition unit;in terms of the software,relevant programs of data acquisition module,preprocessing and feature fusion module,identity registration module were designed.As verified by the experiment,the identification system based on face and voiceprint combination verification has higher accuracy than single biometrics in identification.With better recognition rate and higher safety factor than the single biometric recognition system,the identification system design introduced in the paper has certain practical significance and reference value.