Raw gesture dataset (labelled) which was used for an experimental study aimed at evaluating three machine learning algorithms against several criteria. The data was collected from 10 participants using the MMR+ Mbientlab MetaWear sensor. During data collection, participants were asked to perform 40 repetitions of the 4 selected gestures; circle, swipe right, tap right and tap left. Hence, a total of 160 gestures (4 gestures x 40 repetitions) per participant were collected. Furthermore, participants were initially asked to perform their gestures as consistently as possible, in terms of speed and size of gestures. However, at a later stage in the experiment, participants were instructed to carry out significant variations to the gestures assigned. This included both increasing and decreasing the speed of execution of the specified gestures, as well as pausing briefly during its execution. The repetitions for each gesture were performed in the following manner: - 28 gestures were performed at normal speed. - 4 gestures were performed at a slower speed (in comparison to the gestures performed at normal speed). - 4 gestures were performed at a faster speed (in comparison to the gestures performed at normal speed). - 4 gestures were performed at normal speed but included a brief pause at any point during execution. Each distinct gesture within a participant's data file comprises 40 repetitions, adhering to the specific order as outlined below: the initial 28 files showcase consistent performance, followed by files labeled 29-32 demonstrating a slower execution, then files designated 33-36 showcasing a faster pace, and finally, the last four gestures incorporate a brief pause. The dataset can be used to replicate results but also to conduct other similar studies.