3D Handwriting Characters Recognition with Symbolic-Based Similarity Measure of Gyroscope Signals Embedded in Smart Phone

In this paper, we present a 3D handwriting character recognition algorithm based on a symbolic representation of angular velocity signal generated from Gyroscope sensor included in a smart phone. Characters are writing in a 3D air space by a user via a wearable smart phone device. The main insight behind the use of a symbolic representation is to tackle storage and time processing challenge required in such application with mobile device. That is, we introduce a similarity distance measure that combines the Symbolic Aggregate approXimation (SAX) distance with angular trend distance. The proposed similarity distance measure requires additional information about the trend of the Time Series Data (TSD) and will be used in classification with the nearest neighbor rule. In addition, we extend the distance measure to handle multidimensional data and the classification is achieved using a greedy time warping algorithm. This algorithm internally uses a dynamic programming that overcomes the quadratic time complexity of the classical dynamic time warping search. Experiments on a real database of the 26 lowercase letters in the English alphabet show that our method improves the original and two variation of the SAX similarity search algorithm.