No abstract

No abstract

This research aims at introduction of a hand gesture recognition based system to recognize real time gestures in natural environment and compare patterns with image database for matching of image pairs to trigger unlocking of mobile devices. The efforts made in this direction during past relating to security systems for mobile devices has been a major concern and methods like draw pattern unlock, passcodes, facial and voice recognition technologies have already been employed to a fair level of extent, but these are quiet susceptible to hacks and greater ratio of recognition failure errors (especially in cases of voice and facial). A next step in HMI would be use of fingertip tracking based unlocking mechanism, which would employ minimalistic hardware like webcam or smartphone front camera. Image acquisition through MATLAB is followed up by conversion to grayscale and application of optimal filter for edge detection utilized in different conditions for optimal results in recognizing fingertips up to a precise level of accuracy. Pattern is traced at 60 fps for tracking and tracing and therefore cross referenced with the training image by deployment of neural networks for improved recognition efficiency. Data is registered in real time and device is unlocked at instance when SSIM takes a value above predefined threshold percentage or number. The aforementioned mechanism is employed in applications via user friendly GUI frontend and computational modelling through MATLAB for backend.