Human Action Recognition based on Depth maps, Skeleton and Sensor Images using Deep Learning

Human action recognition has become one of the areas of active research in computer vision for various applications, such as security surveillance, health and human computer interaction. Several approaches for human actions detection are being investigated and images are in RGB (red, green, and blue), depth, and skeleton datasets, as well as inertial sensor images. The majority of the algorithms for action categorization employing skeleton datasets are limited in various ways, After data acquisition for simplicity very basic feature extraction techniques are applied to each data type. The first input is depth images For accuracy of action classification, Neural networks channels are trained with a range of inputs, the second input which is a proposed skeleton images that represents the motion of joints in time, and the third input as inertial images. Neural Networks are taken for evaluation model purpose, then to find Score fusion we are planning to use Avg and Max products. Our proposed method was implementation on public datasets like MAD and UTD-MHAD datasets.