Hand gesture recognition (HGR) has a vital role to develop an intelligent interface for human–computer interaction. This letter proposed an automated system for HGR using three ultrawideband radars. In this system, the received radar reflection is measured for each hand gesture action which is recorded with respect to time duration and distance of hand from radar during hand motion. A time-distance matrix is formed for each gesture action, and reflection density is obtained for time-distance analysis. Furthermore, mean density function-based parameters are suggested, namely, reflection density per unit time, reflection density per unit distance, and mixed moments of time and distance with higher order functions. Features are computed in entire time-distance region and further limited to local time-distance ranges to extract the gesture distinguishable information. Computed features are ranked and given to the classification model. The achieved high accuracy justifies the primacy of features and the developed system.