Vision System for Hand Gesture Recognition (VISOR)
Vision system for Hand Gesture Recognition (VISOR) is a software application that recognizes a set of dynamic continuous gestures (using only a single hand) from a predefined vocabulary using computer vision algorithms. The system is user-independent and able to contend with different backgrounds an...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | text |
| Language: | English |
| Published: |
Animo Repository
2006
|
| Subjects: | |
| Online Access: | https://animorepository.dlsu.edu.ph/etd_bachelors/14182 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | De La Salle University |
| Language: | English |
| Summary: | Vision system for Hand Gesture Recognition (VISOR) is a software application that recognizes a set of dynamic continuous gestures (using only a single hand) from a predefined vocabulary using computer vision algorithms. The system is user-independent and able to contend with different backgrounds and does not require the user to wear a long-sleeved garment (a limitation commonly found in similar systems). A standard USB web camera, placed near the workstation that contains the application, is used to capture gesture sequences. In every frame captured, the hand is detected and afterwards tracked. The hand is detected using a combination of skin-color segmentation and shape analysis based on the concept of convexity defects and several heuristics regarding the human hand shape, established empirically. Upon the hand detection, the user is given a time limit, in which the gesture sequence must be performed. The gesture is recognized by means of extracting temporal and structural information such as motion, hand shape and orientation from the image sequence. In order to recognize the sequence as one of the valid gestures, scoreboarding with respect to the features mentioned is applied after feature extraction.
The system's performance is evaluated in an indoor environment. The prototype was able to detect a user's bare hand and track it throughout the duration of gesture performance. A hand detection rate of 77.65% was achieved with 6.47% false positives, during testing. The system was able to recognize gestures with a success rate of 70% for the old approach and 80% for the new approach, while being able to operate in real-time. |
|---|