Researches On Acoustic-based Handwritten Signature Verification

With the rise and application of information technology,many computer applications require users to perform login operations to authenticate,such as password login.In order to improve user experience and information security,a large number of identity authentication technologies based on human-specific biometrics have been continuously researched and widely used,such as fingerprint recognition and face recognition.Handwritten signatures,as an important behavioral feature of individuals,have been widely used in the identification of user identities in the fields of finance,law,government,etc.,and have a long history.However,a forged signature may cause huge losses and even turbulence.Correct screening is particularly important.Until now,handwritten signature recognition has mainly been performed manually,and the signature verifier determines whether the current signature is a real signature based on the user's historical signature.This method has the disadvantages of high labor costs,reliance on verifier experience,and high rate of false positives.Since the automatic handwritten signature authentication complies with users' demands for user experience and security improvement,it frees human resources from signature recognition and improves the accuracy and stability of signature recognition.Therefore,the research and application of automatic signature authentication systems have been gradually carried out.At present,researches on handwritten signature authentication at home and abroad are mostly based on open datasets.These researches need to be improved in terms of universality and user experience.On the other hand,exploring different signature capture methods can reduce the cost of signature authentication and make signature authentication systems popular in practice.Aiming at the problems existing in the existing research,this paper uses sonic to record signature information and refers to the traditional signature authentication framework to explore a new handwritten signature authentication scheme.The main research results of this article include:(1)It is proposed to use sonic sensing technology to record signature movement.Use the speakers and microphones on mainstream smartphones for sound wave transmission and reception operations,design user-friendly sound wave signals for the characteristics of commercial hardware,and use sound wave phase-related information to track small movements in handwritten signature movements.(2)A suitable feature extraction and user-independent recognition model is designed for the acoustic wave phase related information.The environmental noise in the phaserelated information is removed,and the information of the handwritten signature action is extracted.Combining with the mainstream automated signature authentication system framework and classification technology,a handwritten signature discrimination process and model suitable for acoustic waves are designed.(3)Perform experimental evaluation and prototype system implementation.Comprehensive evaluation of the scheme in this paper,including verification accuracy,cross-user availability,effects of hardware compensation,system robustness,microbenchmarking Experiments,comparative experiments,and replay attacks.Designed and implemented a sonic-based signature authentication prototype system and evaluated its operating efficiency.The system's accuracy performance reached: AUC(area under the ROC curve)= 98.7% and EER(equal error rate)= 5.5%.Experimental results shows that the scheme is a non-intrusive,robust,secure,low-latency handwritten signature authentication scheme.This article first introduces the background and research significance of the handwritten signature authentication system,analyzes the current status and shortcomings of the current research,and then briefly summarizes the handwritten signature authentication system and proposes the technical route of this article.Based on this,a sonic-based handwritten signature authentication scheme with four modules is researched and designed,and key technologies such as sonic phase-dependent signal sensing technology,feature extraction technology,similarity measurement technology,and deep learning models are studied in depth.The design ideas and technical details of each module in the scheme are introduced in detail.Then,several schemes of experiments were designed to evaluate the scheme,and a prototype system was designed and implemented.Finally,this article summarizes the research content and looks forward to future research work.