Research Of Face Recognition Technology For Identity Verification On Mobile Devices

In recent years, computer technology and mobile devices such as smart mobilephone etc. become popular and obtain highly development. The network securityproblem with effective and convenient of mobile equipment during the identityauthentication process is gradually highlighted its importance. Since identificationtechnology based on biological characteristics has many security advantages such asdifficult to be copy, forged, it has gotten widely attention in both academia andindustry. In these biological characteristics, e.g. fingerprint, retina, gait, and so on,face recognition technology is the most rapid and friendly way of identifying in dailylife, at the same time it is also a research hotspot in the field of pattern recognition.With the rapid development of3G network and the enhancement of mobile hardwareconfiguration, the realization of the identity authentication technology on mobileterminal equipment becomes feasible. However, the existing research on facerecognition technology for mobile equipment is still uninvestigated. Traditional facerecognition technology is ineffective when the illumination or background of mobiledevices changed, and the posture, eyes, facial expressions of users varied. In thispaper, we take the inherent advantages of Android platform to research facerecognition technology on mobile platform. The main works are given as follows:First, propose a high accuracy face detection technology. The main challenge forface detection systems usually include complex background, changes in posture,facial expression variety, image positioning unreasonable, poor imaging conditionsetc.. Aiming at these problems, we proposed a color image edge information and skincolor information based on a face detection algorithm which is not sensitive toillumination. The algorithm firstly uses image enhancement preprocessing operationto adjust input image illumination, and then combines the skin color method and theedge information to improve the speed and accuracy of face detection, adopts theoriginal image feature to improve face authentication effectively. The algorithm candetect images of different size, position conditions, and there are not constraints onthe illumination and facial expressions.Second, develop a face recognition technology on memory constrained mobiledevices. For the problem of the limited memory and computing resources on mobiledevices, we develop an algorithm based on identification information value andnearest neighbor classifier algorithm. The algorithm guides the allocation for limited memory resources according to the characteristics of stream data classificationprocessing to maximize the recognition performance, and optimized the subsetidentifies by previously captured faces to obtain best choice subset to save in memory.The algorithm has considered identify potential value of each data, estimated thestored classifying collection value, shown the expected information in system. Inaddition, the algorithm also provides the relevant principle of dynamic data set in thelimited memory condition.Third, according to the above four research contents, we realize a fast andefficient identification system for mobile terminal. The functions of the identityauthentication system include: data collection function model, including userregistration function and image collection function; pretreatment model, includingface detection and face recognition training function; recognition model, include facerecognition function and password authentication. The system use the C/S modelsystem structure, combined with the Android system platform, ASP.NET developmentplatform to realize each function model.