Multichannel Symbolic Aggregate Approximation Intelligent Icons: Application for Activity Recognition

In this work, we introduce the Multichannel Intelligent Icons, a novel method for producing and presenting essential patterns of multidimensional bio-signals. The proposed approach is an extension of Symbolic Aggregate Approximation (SAX) along with an innovative variation of Intelligent Icons. The innovation on the approach stands on the grounds of creating a spatial correlation of the inherited information in all dimensions and so it provides extra features for distinguishing the human activities. The proposed model is testing on Human Activity recorded data and for the classification purposes a Nearest Neighbour classifier is applied. The achieved results are compared with the case of applying single-channel intelligent icons approach and it is inferred a noteworthy increase in terms of accuracy and sensitivity with the proposed approach.