Overview:
In this project, we designed a Signature Authentication Pen, which realizes identity authentication by exploiting the signature biometric features of the users. Our Signature Authentication Pen collects the acceleration and angular velocity data when the user is writing, and build machine learning model for real time identity prediction. As the biometric features of writing is hard to imitate, our system is much safer compared with traditional identity authentication methods. Our project provides a new approach for identity authentication and has a broad application prospects in shopping, criminals recognition and smart home.
Our project works in two modes: real time identity recognition and fake signature distinguishing.
In the real time identity recognition mode, different users write the same word, our signature authentication pen can identify who write the word. In this mode, the test user must belong to one of the users in the training classes. The output of test result is the class of which the user belongs to.
In fake signature distinguishing mode, after a test user write the name, our signature authentication pen can distinguish whether it is the fake signature. In this mode, the training set contains the real signature of that user and a group of fake signatures. The output of the result is true or false to indicate whether the signature is real. Moreover, the test user needn’t be inside the training set.
Our project works in two modes: real time identity recognition and fake signature distinguishing.
In the real time identity recognition mode, different users write the same word, our signature authentication pen can identify who write the word. In this mode, the test user must belong to one of the users in the training classes. The output of test result is the class of which the user belongs to.
In fake signature distinguishing mode, after a test user write the name, our signature authentication pen can distinguish whether it is the fake signature. In this mode, the training set contains the real signature of that user and a group of fake signatures. The output of the result is true or false to indicate whether the signature is real. Moreover, the test user needn’t be inside the training set.
Equipment:
Grove - Universal 4 Pin Buckled 50cm Cable (5 PCs Pack)
MetaWearC Streaming Sensor - bluetooth wireless sensor
Customed pen
Grove -3-Axis Digital Accelerometer
Grove - 3-Axis Digital Gyro
Intel Edison