Informatics is becoming essential in our daily activities. The use of the human machine interface requires fine movements from the user. When human movements are distorted, for instance, by tremor, performance could be improved by digitally filtering the intermediate signal before it reaches the controlled interface. This paper presents a novel device able to filter out tremorous movements from the cursor of a mouse in the screen of the computer. This task requires real-time discrimination between voluntary and tremorous movement. To address this problem, a learning algorithm is introduced as well as the evaluation and validation with users of the prototype developed.