**2. Related Work**

Scientists have employed various approaches to implement FDSs over the past years. They have been classified as presented in Figure 1. Each of them has its strengths and weaknesses. We focus on wearable technologies since we use this approach. Nevertheless, several survey studies [9,10] reported the other methods in more depth.

### *2.1. Choice of Sensors and Sampling Rate*

Several types of sensors including accelerometers, gyroscopes, magnetometers and tilt sensors have been used to detect falls. Based on the fall characteristics, most studies, such as [11–15], employed only acceleration measurements. From our literature review, very few studies use a single gyroscope. For example, Bourke and Lyons [16] used a single biaxial gyroscope and measured changes in angular velocity, angular acceleration and body angle. Tang and Ou [17] also reported promising results, using a single six-axis gyroscope. The separate use of these sensors already produced promising results but their combination is even better [18]. Wang et al. [19] employed a heart rate monitor and discovered that the heart rate increases by 22% after a fall in people over 40 years old. This demonstrates that physiological data can be used in such a system. Across the papers reviewed (summarized in Table 1), the sensors' sampling rate varied within a range from 10 to 1000 Hz. This variation is not small, one having 100 times more samples than the other, seemingly arbitrarily. Fudickar et al. [20] compared the detection results when varying the sampling rates from 50 to 800 Hz. The results obtained with a sampling rate of 50 Hz were as good as the ones with 800 Hz. Other studies show that low sampling rates can offer reasonable results, for example Medrano et al. [21] used data sampled up to 52 Hz. We further investigate this issue in this paper with similarly low sampling rates.

**Figure 1.** Classification of Fall Detection System approaches.

**Table 1.** Reviewed studies that used wearable sensors for fall detection (including acronyms at the end of the table).



BDM: Bayesian Decision Making; DTW: Dynamic Time Warping; LDA: Linear Discriminant Analysis; LSM: Least Squares Method; PPCA: Probabilistic Principal Component Analysis; WT: Wavelet Transform; ANN: Artificial Neural Network; MLP: Multilayer Perceptron; RBF: Radial Basis Function.
