基于人脸识别的智能推送服务取号系统研究

发布时间：2018-03-18 01:37

本文选题：人脸识别　切入点：服务推送　出处：《海南大学》2017年硕士论文　论文类型：学位论文

【摘要】：传统的Web推送服务大多是根据系统的登录账号、历史记录或者订阅记录的分析判断进行内容的推送,这样的推送机制不能充分体现使用者个体间的喜好差异。自从Web2.0技术推出以来,基于该技术的智能化、个性化的推送服务和应用成为该领域的研究热点。本文根据服务大厅的个性化服务的实际需求,探讨将人脸识别技术与服务推送技术相结合,实现基于人脸识别的个性化、智能服务推送系统。在实际的推送服务系统中,一个用户在样本库中通常只有一幅头像,所以本文将重点探讨单样本的人脸特征提取算法和识别技术。在分析基于主成份分析法的单样本人脸识别经典算法((PC)2 A、SPCA)优缺点的基础上,提出了通过SPCA算法提取全局特征,通过M(2D)2PCA+PCA算法提取局部特征,然后通过最大隶属度分别求得基于全局特征和基于局部特征的识别结果,最后采用模糊综合方法在决策层进行加权融合,并在ORL人脸库和Yale人脸库上进行了算法验证。实验结果表明,该算法具有良好的识别性能,在ORL人脸库的识别率为89.94%,在Yale人脸库的识别率为90.94%;并且对表情、姿态、光照等的变化具有较好的鲁棒性。为了提供更精准的个性化服务推送,对基于用户个人数据分析的个性化服务技术进行了深入的探讨,建立了基于用户协同过滤算法(UBCF)模型,最后通过PUSHLET技术将服务信息推送给用户。最后,以服务大厅取号机系统为例,研发了基于人脸识别的智能推送服务取号原型系统,该系统由人脸识别、账号管理、信息获取以及服务推送等四部分构成,实现了服务推送系统的个性化、智能化的目的。
[Abstract]:The traditional Web push service is mostly based on the system's login account, history or subscription records of the analysis of the content of the push, This push mechanism does not fully reflect the differences of preferences among users. Since the introduction of Web2.0 technology, based on the intelligence of that technology, Personalized push service and application have become the research hotspot in this field. According to the actual demand of personalized service in service hall, this paper discusses the combination of face recognition technology and service push technology to realize personalization based on face recognition. An intelligent service push system. In a real push service system, a user usually has only one image in the sample library. So this paper will focus on the single sample face feature extraction algorithm and recognition technology. On the basis of analyzing the advantages and disadvantages of single sample face recognition classic algorithm based on principal component analysis, this paper proposes to extract global features by SPCA algorithm. The local feature is extracted by MJ 2DX 2PCA PCA algorithm, then the recognition results based on global feature and local feature are obtained by the maximum membership degree. Finally, the fuzzy synthesis method is used for weighted fusion at the decision level. The algorithm is validated on ORL face database and Yale face database. The experimental results show that the algorithm has good recognition performance, the recognition rate in ORL face database is 89.94, the recognition rate in Yale face database is 90.94, and the recognition rate is 90.94 for expression and pose. In order to provide more accurate personalized service push, the personalized service technology based on user personal data analysis is deeply discussed, and the UBCF-based user collaborative filtering algorithm is established. Finally, the service information is pushed to the user by PUSHLET technology. Finally, a prototype system of intelligent push service number retrieval based on face recognition is developed, which is managed by face recognition and account number. Information acquisition and service push are composed of four parts, which realize the purpose of individuation and intelligence of service push system.
【学位授予单位】：海南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP391.3

【相似文献】