Hand Body Language Gesture Recognition Based on Signals From Specialized Glove and Machine Learning Algorithms

The man-machine interface (MMI) is one of the most exciting areas of contemporary research. To make the MMI as convenient for a human as possible, it is desirable that efficient algorithms for recognizing body language are developed. This paper presents a system for quick and effective recognition of gestures of hand body language, based on data from a specialized glove equipped with ten sensors. In the experiment, 10 people performed 22 hand body language gestures. Each of the 22 gestures was executed 10 times. Collected data were preprocessed in multiple ways and three machine learning algorithms were designed based on classifiers (probabilistic neural network, support vector machine, and k-nearest neighbors algorithm) trained and tested by a tenfold cross-validation technique. The best designed classifiers gained effectiveness of gesture recognition at κ = 98.24% with a very short time of testing, below 1 ms. The experiments confirm that efficient and quick recognition of hand body language is possible.