Kertas kerja ini menerangkan tentang pembangunan satu sistem pengesahan tandatangan bertulis tangan yang melibatkan tekanan pen terhadap laluan tandatangan, masa ketika menandatangan, profil kelajuan dan kedudukan rupa bentuk tandatangan. Isyarat bertulis tangan telah diperoleh dan diolah secara berdigit menggunakan tablet. Ciri utama sistem pengesahan tandatangan yang dicadangkan ialah tandatangan bertulis tangan yang dikemaskini secara dinamik, keupayaan cuba semula semasa pengesahan, kegunaan jalur terima beserta nilai ambang, pembangunan mesra pengguna berdasarkan antara muka grafik pengguna, penggunaan kaedah paksi masa sepunya dan pengesahan tandatangan menggunakan satu kelas rangkaian neural laluan hadapan berlapis. Satu algoritma khusus telah diguna pakai yang dapat memberikan keputusan pengesahan dengan ketepatan yang baik serta lebih cepat. Sistem telah menghasilkan kadar penolakan palsu sebesar 1.3% dan kadar penerimaan palsu 0% dengan pengesahan dilakukan menggunakan tandatangan palsu yang telah diciplak.
Kata kunci: Biometrik, penentusahan tandatangan, perolehan data, jalur terima, rangkaian neural
The paper describes the development of a handwritten signature verification system incorporating pen pressure of signature path, time duration of the signing procedure, velocity profile of signature and position of signature shape. The handwritten signals have been captured and digitized using a tablet. The main features of the proposed signature verification system are the dynamically update of handwritten signature, retries capability in verification, application of tolerance bands and threshold values, development of user friendly Graphic User Interface, application of Common Time Axes and verification of signatures using a class of a multilayer feed-forward neural network. A novel algorithm has been applied that provides the ability to produce consistent and high accuracy verification result and maintain the speed of verification. The system has yielded 1.33% of False Reject Rate and 0% False Acceptation Rate with the verification using random forgery signatures.
Key words: Biometrics, signature verification, data acquisition, tolerance bands, neural network
Online handwritten signature verification system based on DWT features extraction and neural network classification
