A Combined Method of Skin-and Depth-based Hand Gesture Recognition

Abstract

Kinect is a promising acquisition device that provides useful information on a scene through color and depth data. There has been a keen interest in utilizing Kinect in many computer vision areas such as gesture recognition. Given the advantages that Kinect provides, hand gesture recognition can be deployed efficiently with minor drawbacks. This paper proposes a simple and yet efficient way of hand gesture recognition via segmenting a hand region from both color and depth data acquired by Kinect v1. The Inception model of the image recognition system is used to check the reliability of the proposed method. Experimental results are derived from a sample dataset of Microsoft Kinect hand acquisitions. Under the appropriate conditions, it is possible to achieve high accuracy in close to real time

Two viewpoints based real‐time recognition for hand gestures

“…Finger count and finger position relative to palm are prominent features in the correct recognition of gestures [27]. Fingers can be detected in many ways such as convexity defects [26], K ‐cosine corner detection [20], and colour depth [10]. The finger count and detection method encompasses calculating the centre of gravity from the segmented hand and thereafter determining the maximum distances to tangent points of fingers [9].…”

“…Positions of thumb vividly contrast between certain characters [28]. Scale‐space feature detection can find palm‐like and finger‐like structures, so that the hand gesture is determined by the palm finger configuration [10, 19]. Curvature measure for blobs is used to determine the fingertips [10, 11] in images acquired from Alice.…”

“…Scale‐space feature detection can find palm‐like and finger‐like structures, so that the hand gesture is determined by the palm finger configuration [10, 19]. Curvature measure for blobs is used to determine the fingertips [10, 11] in images acquired from Alice. Our method of determining finger position is generally by finding the finger width to palm width ratio.…”

“…Favourable illumination will result in a more informative and favourable processed image. Handshape [1, 3, 8, 9] and finger information [3, 9–11] are major features which contribute to hand gesture recognition. The approach to realise these features varies based on the nature of gesture and its environment.…”

Using VGG Models with Intermediate Layer Feature Maps for Static Hand Gesture Recognition

“…However, they implemented the model using only six classes of hand gestures. Other researchers 15 used the inception model for hand gesture recognition, in which they fed the model with depth image data in addition to the color image data acquired by an acquisition device known as Kinect. Using transfer learning to train the last layer of the model on a target dataset of 10 classes, they reported highly accurate results for their model.…”