<![CDATA[Signature is a someone unique identity used in presence system as attendance proof and official documents approval. Besides written on paper, signatures are also recorded digitally. To recognize signatures in the realm of technology is called biometrics while the method often used is backpropagation. Signatures are included in the biometric behavioral category with one of the dynamic characteristics of measurement carried out by calculating duration of writing and changes in writing movements. One of the motion recorder technologies is MPU6050 sensor. This sensor is used to get the writing duration feature through many data sets recorded when signature, then changes in vertical movement through gx and changes in horizontal movement through gy. This feature is extracted to be calculated by backpropagation method in tensorflow and requires training data and test data. The training data of this study uses signatures of five different people with five signatures for each person. Testing is done by predicting the five signatures of five people whose signature patterns have been recorded. From the MPU6050 sensor testing, MAPE 17% is obtained. On signature recognition using backpropagation, MAPE was 33%. On prediction signature shows an accuracy of 80%. And at the time of computing requires approximately 54.63s.]]>
Copyrights © 2020
