Smart TV Control System Using Gesture Recognition Based On Single-camera

Gesture, as a natural way for human computer interface, plays a vital role in theinteraction with smart TV. However, related researches about gesture based interaction systemfor smart TV are faced with enormous challenges, including hand tracking as a non-rigid andfast-move object, unpredictable light ambient and background in home environment, as wellas the implementation of a real-time, low-cost system.This paper designs and develops a gesture based interaction system based on a webcamera, especially for创维42E79U smart TV. The main framework could be divided intothree stages: hand detection, hand tracking and gesture recognition. The chief work of thispaper includes the following three parts:In hand tracking, this paper makes an improvement upon incremental PCA algorithm forvisual tracking. In order to address the complex ambient factors mentioned above, theappearance model could updates the subspaces adaptively. Meanwhile, incremental weightedPCA algorithm could ease the undesirable influence from the background. Moreover, thepaper proposes the double sampling and movement prediction based on particle filtering,which could capture the hand in a fast movement. Finally, the experiment demonstrates theeffectiveness of the methods proposed in this paper.In gesture recognition, this paper conducts a detailed analysis about the interactionscheme of the smart TV, designs a simple but effective gesture instruction set. For this set, thispaper applies the Levenshtein distance to measure the similarity for the recognition. With thepre-processing of the trajectory, effectiveness checking and variance based similarityclassification, this paper achieves a precision of0.96in motion gesture recognition.At last, in system design, this paper proposes a server-client structure. The complexalgorithm running load is assigned to the server, which is a high performance PC. Meanwhile,the response of the instruction is displayed in the client, that is, the smart TV. The server andthe client are connected by the Internet. The experiment also demonstrates that, compared tothe built-in hand control system in the smart TV, the system designed in this paper is moresuitable for practical use, and maintains a better attribution of compatibility and extension.