QRD RLS Algorithm for Hand Gesture Recognition Applications

The paper is focused on an algorithmic technique for detection of hand presence and distance from a hand to device transmitting ultrasound signals. The described method is based on a QRD Recursive Least Squares (RLS) algorithm with double precision arithmetic and exponential forgetting (EF). Modelling of a hand detection problem is based on linear Finite Impulse Response (FIR) based regression models and performed using MATLAB tools. The modelled system comprises an environment model, a hand model and an identification block. A series of experiments testing both time-invariant and time-variant environment models and time-variant hand models show the importance of a correct choice of the EF factor. The experiments have proven the accuracy of the algorithm and the possibility to calculate a distance from the hand to the device. The final version of the algorithm is supposed to be implemented on the embedded Xilinx Zynq device equipped with a microphone and ultrasound transducers.