Development of Cost-Efficient and Accurate Tactile Sensing Wristband for Real-Time Static Gesture Recognition
Paulus, Aliisa (2020)
Paulus, Aliisa
2020
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2020120225686
https://urn.fi/URN:NBN:fi:amk-2020120225686
Tiivistelmä
There has been a tremendous increase in the use of gestures in human-computer interaction technology, where the most common approach is vision-based systems. However, a tactile sensing device around the wrist can give more information about a hand's static gestures than only motion tracking. The sensors can detect the pressure distribution on a surface between the sensor and the wrist. Hence, the thesis aimed to design and create a tactile sensing wristband prototype for a real-time hand gesture recognition system.
An in-depth analysis of current gesture recognition systems was conducted, from which a suitable architecture and machine learning algorithm was identified. After carefully selecting the appropriate components and techniques for the wristband, the final prototype was designed, constructed, and tested to obtain its performance. The main focus lies in the classification of the pressure values into gestures with the Support Vector Machine algorithms. The sensor values were scaled with a Min-Max normalization and then used as input for the Support Vector Machine, which included a radial base function kernel.
The result showed that all ten static gestures were clearly distinguished. The classification accuracy for the static gestures in the experiment was 98.0%. The result obtained shows that the wrist-based gesture recognition prototype adhered to the requirements presented in this thesis.
An in-depth analysis of current gesture recognition systems was conducted, from which a suitable architecture and machine learning algorithm was identified. After carefully selecting the appropriate components and techniques for the wristband, the final prototype was designed, constructed, and tested to obtain its performance. The main focus lies in the classification of the pressure values into gestures with the Support Vector Machine algorithms. The sensor values were scaled with a Min-Max normalization and then used as input for the Support Vector Machine, which included a radial base function kernel.
The result showed that all ten static gestures were clearly distinguished. The classification accuracy for the static gestures in the experiment was 98.0%. The result obtained shows that the wrist-based gesture recognition prototype adhered to the requirements presented in this thesis.