3.1.1. Smart-Band Dataset

The smart-band dataset created by Yanay et al. contains air-writing data collected from 55 subjects [2]. Each subject wrote 15 sets of all 26 letters in the English alphabet in the air wearing a smart-band on their wrist. So, the dataset contains 390 samples of air-written letters per subject and there are 21,450 samples in total in the dataset.

The subjects wore the ‘Microsoft Band 2’ smart-band. The smart-band motion sensors, e.g., accelerometer and gyroscope, were used to record motion measurements in each sample. Both the accelerometer and gyroscope data were taken from three different axes (X, Y, and Z) with a maximum sampling rate of 62 Hz. To collect the signals from the motion sensors, an android mobile application was developed which ran on a Samsung Galaxy S8 smartphone. The smart-band was worn in the wrist of the hand with which the subject normally wrote, and the smartphone was held with another hand. The smartphone was connected to the smart-band via Bluetooth to collect data. Among the 55 subjects, 28 were females and the rest were males, while 46 subjects were right-handed, and the remaining 9 subjects were left-handed.