*3.1. Dataset*

We decided to use a publicly available dataset rather than creating our own experiment with diverse subjects, for reproducibilty purposes. Therefore, in order to select such dataset, we considered those evaluated in a recent meta-review [39]. From those, we preselected those who were freely available, as listed in Table 2 which describes each dataset characteristics. We decided to use acceleration measures because studies have shown that interesting performances can be achieved with it. Ultimately, we selected the dataset named *SisFall* [31] over others [40,41] because of its high quality. We assessed this quality with various criteria, namely the size of the dataset and the diversity of subjects in terms of age, gender, weight and height, as detailed in Table 2.

**Table 2.** Main characteristics of the considered datasets. Adapted from [39].


A: Accelerometer; G: Gyroscope; M: Magnetometer; An: Ankle; Ch: Chest; Th: Thigh; Wa: Waist; Wr: Wrist.

We also took into account the number of falls and ADLs performed by each subject. An additional factor was the sensors' sampling rate which needed to be high in order to experiment using various sampling rates. In the *SisFall* dataset, two tri-axial accelerometers (ADXL345 and MMA8451Q) and a tri-axial gyroscope (ITG3200) were used at a sampling rate of 200 Hz. These sensors were attached to the waist, following the longitudinal axis, of the subjects in the data collection phase [31]. This location has been proven to be a reliable one from the literature, as discussed in Section 2.2.

We decided not to use the data of the second accelerometer (MMA8451Q) because usual setups only have a single accelerometer. Having decided to use only data from one accelerometer, we chose that with the highest sensing range and the lowest power consumption which seems adequate for the application. Future work could explore whether there is a significant difference between these sensors.

Twenty-three young people (19 to 30 years old) performed 15 types of falls and 19 types of ADLs including fall-like activities. Fifteen elderly people (60 to 75 years old) also performed the same ADLs for more authenticity. There were five trials per activity except for the walking and jogging activities, each of which had only one trial (See Table 3). Hence, *SisFall* contains a total of 4505 records including 2707 ADLs and 1798 falls, making it unbalanced. A total of 38 people including 19 women and 19 men participated. Table 3 lists the falls and ADLs and their duration.


**Table 3.** Details of the Activities of Daily Living and falls contained in the *SisFall* dataset [31].

