The following code allows to recognize an action which has been triggered by voice using a virtual assistant. Nowdays there are plenty assistant which allows you to perform same task like increase or lower the volume or stop and play a track. In this project we had to recognize 7 commands, each of them was recorded in our dataset and was taken from a real case scenario. Using Mel Frequency Spectrum we were able to reach a 92% accouracy without using neural network, we had the best performances using SVM with an RBF kernel. For more details about the process please check DSL_project_report