Chao Xu, Parth H. Pathak, et al. HotMobile’15 Finger-writing with Smartwatch: A Case for Finger and Hand Gesture Recognition using Smartwatch Chao Xu, Parth H. Pathak, et al. HotMobile’15
What They Did Classify gesture type: Finger, Hand or Arm Recognize 37 gestures Recognize finger writing with smartwatch
What They Did Classify gesture type: Finger, Hand or Arm Recognize 37 gestures Recognize finger writing with smartwatch
Experiment and Observation
Methodology Data Collection Feature extraction Classification Use Shimmer device to collection data Data from accelerometer sensor and gyroscope sensor at 128 Hz Feature extraction Formula Magnitude values are FFT coefficients calculated over the time window Classification
What They Did Classify gesture type: Finger, Hand or Arm Recognize 37 gestures Recognize finger writing with smartwatch
Gesture Types & Theory Gesture Types Recognize Theory
Feature Extraction & Evaluation Features Used for Classification Information Gain on Each Feature
Identification Performance Gesture Recognition Accuracy TP rate of Naïve-Bayes Classifier
What They Did Classify gesture type: Finger, Hand or Arm Recognize 37 gestures Recognize finger writing with smartwatch
Data Collection The size of the alphabet is 2.5” in width and height Accelerometer and gyroscope data is collected Each of the 26 alphabets is repeated 10 times The features are the same as those used in gesture recognition
Performance Logistic regression outperforms the other two classifiers in accuracy Average accuracy is 94.6%
Discussion This paper reveal the potential that the smartwatch can be used to detect fine-grained movements of user’s fingers There are still challenges to realizing the true potential of smartwatch In experiments, the user’s wrist and arm are affixed to the chair arm Different people write and perform different gestures in different ways Finger-writing in the air Detecting continuous writing to form words and sentences