technologies

Article
Deep Learning Based Fall Detection Algorithms for
Embedded Systems, Smartwatches, and IoT Devices
Using Accelerometers
Dimitri Kraft 1, * , Karthik Srinivasan 2 and Gerald Bieber 3, *
1 Faculty of Computer Science and Electrical Engineering, University of Rostock, 18059 Rostock, Germany
2 Next Step Dynamics AB, 211 19 Malmö, Sweden; [email protected]
3 Fraunhofer-Institut fuer Graphische Datenverarbeitung IGD, 18057 Rostock, Germany
* Correspondence: [email protected] (D.K.); [email protected] (G.B.)




Received: 30 October 2020; Accepted: 29 November 2020; Published: 2 December 2020


Abstract: A fall of an elderly person often leads to serious injuries or even death. Many falls
occur in the home environment and remain unrecognized. Therefore, a reliable fall detection is
absolutely necessary for a fast help. Wrist-worn accelerometer based fall detection systems are
developed, but the accuracy and precision are not standardized, comparable, or sometimes even
known. In this work, we present an overview about existing public databases with sensor based
fall datasets and harmonize existing wrist-worn datasets for a broader and robust evaluation.
Furthermore, we are analyzing the current possible recognition rate of fall detection using deep
learning algorithms for mobile and embedded systems. The presented results and databases can
be used for further research and optimizations in order to increase the recognition rate to enhance
the independent life of the elderly. Furthermore, we give an outlook for a convenient application and
wrist device.

Keywords: fall detection; accelerometer; datasets; deep learning; neural networks; wrist; smart bands;
watches; IoT devices; edge computing

1. Introduction
The independent life of an elderly person can be changed drastically after a fall. Depending on
the health condition of the elderly, almost 10 percent of the people who fall will suffer from serious
injuries, or might even die directly after a fall if no intermediate help is available [1]. To prevent
the severe consequences of such falls, a reliable fall detection is needed. One common approach to fall
detection is using wrist worn detection systems that are measuring acceleration forces. These wrist
devices are gaining more and more acceptance across the population and becoming increasingly
powerful in terms of computational performance that the usage of artificial intelligence is reasonable.
In general, older adults appear to be interested in using such devices although they express concerns
over privacy and understanding exactly what the device is doing at specific times [2]. The evaluation
of mobile fall detection systems is highly sophisticated because live data from falls of elderly people
are rare. Boyle et al. tried to use real-time data with 15 adults over the course of 300 days and was only
able to record four falls during that time [3]. Even simulated data are barely available and they are
existing only in various characteristics. Therefore, the aim of this paper is

• to present existing public datasets of fall detection,

• to describe the harmonization process for these datasets,
• to state the current accuracy of fall detection for tiny, mobile and embedded systems using deep
learning algorithms,

Technologies 2020, 8, 72; doi:10.3390/technologies8040072 www.mdpi.com/journal/technologies

Technologies 2020, 8, 72 2 of 17

• to invite and motivate researchers to compete and contribute in fall detection by using the existing
databases and to provide various achievements for the future life of the elderly.

2. Related Work
The detection of falls can be done through several approaches and technologies. Some approaches
are infrastructure based and using external cameras [4] or floor sensors [5,6], while other approaches
consist of mobile sensor analysis through body worn sensors, e.g., accelerometers, gyroscopes, and air
pressure sensors. Each system records a sensor data stream that can be analyzed to recognize
a fall [2]. To distinguish between no fall and fall events, various strategies were employed. Often,
datasets are used to train machine learning algorithms. Other approaches do not rely on certain
datasets as they consist of constructing rules to distinguish between those events. The quality of such
datasets is an important constraint for the quality of the trained fall detection algorithm.

2.1. Datasets
The availability of fall data is mandatory for the development of fall recognition systems. At first,
datasets were interspersed throughout the scientific community until Casilari et al. [7] provided
a comprehensive overview of publicly available fall detection datasets. Since then, new datasets were
published. We augmented the overview of Casilari et al. by adding recently published datasets to our
overview table (see Table 1) with their corresponding characteristics (see Table 2).

Table 1. Overview of fall detection datasets recorded with a body worn sensor.

Sensor
Dataset Reference Sensors Sampling Rate Year
Location
DLR [8] A, G, M Waist 100 Hz 2010
MobiFall [9] A, G, O Thigh 87 Hz (A) 100 Hz (G, O) 2013
Waist
TST Fall Detection [10] A 100 Hz 2014
Right Wrist
Thigh
tFall [11] A 45 Hz 2014
Hand bag
UR Fall Detection [12] A Waist 256 Hz 2014
Head, Chest,
Simulated Falls
[13] A, G, M, O Waist, Right Ankle, 25 Hz 2014
and ADL
Right Wrist, Right Thigh
Cogent Labs [14] A, G Chest, Thigh 100 Hz 2015
Project Gravity [15] A Thigh, Wrist* 50 Hz 2015
Graz [16] A, O Waist 5 Hz 2015

Ankle, Chest,
MUMAFall [7] A, G, M 100 Hz / 20 Hz* 2016
Thigh, Waist, Wrist

Right Ankles, Left Ankles,

SFU Fall Detection [17] A Right Thighs, Left Thighs, 128 Hz 2016
Waist, Sternum, Head
SisFall [18] A1 , A2 , G Waist 200 Hz 2017
UniMiB SHAR [19] A Thigh 200 Hz 2017
Right Wrist,
SmartWatch [20] A 31.25 Hz 2018
Left Wrist
Right Wrist,
Notch [20] A 31.25 Hz 2018
Left Wrist
CGU-BES [21] A Chest 200 Hz 2018
Waist,Wrist,Neck,
UP Fall Detection [22] A, G 18 Hz 2019
Thigh, Ankle
MARG [23] A Waist 50 Hz 2019
Right Wrist,
SmartFall [24] A 31.25 Hz 2019
Left Wrist
Note: Ai : Accelerometer, G: Gyroscope, M: Magnetometer, O: Orientation. * Some samples are recorded with 20 Hz
while other are recorded with 100 Hz.
Technologies 2020, 8, 72 3 of 17

Table 2. Dataset Characteristics.

Data-Set # Types of ADLS/Falls # Samples of ADL/Falls Duration of Samples (s)

Min-Max Mean Median
DLR 15/1 1017 (961/56) [0.27–864.33] s 18.40 s 9.46 s
MobiFall 9/4 630 (342/288) [0.27–864.33] s 18.40 s 9.46 s
MobiAct 9/4 2526 (1879/647) [4.89–300.01] s 22.35 s 9.85 s
TST Fall Detection 4/4 264 (132/132) [3.84–18.34] s 8.6 s 8.02 s
tFall not typified/8 10909 (9883/1026) 6s 6s 6s
UR Fall Detection 5/4 70 (40/30) [2.11–13.57] s 5.95 s 5.27 s
Sim Fall and ADLS 16/20 3184 ( 1120/1400) [8.44–35.24] s 20.15 s 19.2 s
Cogent Labs 8/6 1968 (1520/448 ) [0.53–55.73] s 13.15 s 12.79 s
Graz 8/6 1968 (1520/448 ) [0.53–55.73] s 13.15 s 12.79 s
MUMAFall 8/3 531 (322/209) 15 s 15 s 15 s
SisFall 19/15 4505 (2707/1798) [9.99–179.99] s 17.6 s 14.99 s
UniMiB SHAR 9/8 7013 (5314/1699) 1s 1s 1s
SmartFall 4/4 2331 (1804/527) N/A N/A N/A
SmartWatch 4/4 181 (90/91) N/A N/A N/A
Notch 7/4 2563 (2456/107) N/A N/A N/A
UP Fall Detection 6/5 578 (306/272) [10–60] s 30 s 10 s
SFU Fall Detection 12/5 600 (240/360) 15 s 15 s 15 s
CGU-BES 8/4 180 (120/60) 16.5 s 16.5 s 16.5 s
MARG 5/13 432 (120/312) [1.3–11.5] s 5.3 s 5.58 s

2.2. Position of Sensors

More than half (10 out of 19) of the datasets described in Table 1 contain waist-worn acceleration
data, while only seven datasets contain wrist-worn acceleration data. We do not focus on thigh data
sets due to the limited practicability in daily life. We suspect that a wrist or waist worn acceleration
sensor is much easier to use and provide a higher wear comfort. Pannurat et al. pointed out that most
fall detection systems using sensors mounted to the torso [25]. Krupizer et al. recently provided
an overview of the most common positions for wearable sensors in fall detection systems [26].
They confirmed that the waist is a popular position for most automatic fall detection systems.

2.2.1. Wrist vs. Waist

Liao al. showed in [27] that their approach performed best on the human activity recognition task
when the data are obtained by a waist worn accelerometer. In this paper, we focus on wrist-worn data,
as we strongly believe that sensor solutions for the wrist are more convenient for the end user and
may be adopted earlier. Furthermore, a fall detection algorithm may be easily integrated into currently
available smartwatch or fitness tracker products.

2.3. Fall Detection Algorithms

Several fall detection algorithms were proposed in the last decade. These algorithms
ranging from simple threshold based approaches [28,29], over handcrafted feature based machine
learning algorithms [30,31] and finally to deep learning based automatic feature extraction neural
networks [24,32,33]. Most of the aforementioned contributed datasets come with a detailed approach for
detecting falls. Some of the relevant work may be found in Table 3. Note that the performance reported
may not express the true capability of these approaches, as the validation methods of these algorithms
are mixing users in their train and validation sets, resulting in a fragile evaluation process.
Technologies 2020, 8, 72 4 of 17

Table 3. Overview about published fall detection algorithms.

Falls/ Validation
Source Dataset Features Classifier Performance
ADLS [n/n] Method

Naive Bayes,
342/ 75–25 split,
[31] MobiFall Time domain LS, ANN, 87.5% accuracy
288 mixed user
SVM

346/ Time and frequency MLP, SVM,

[34] MobiFall 346 test samples 97.29% sensitivity
11080 domain KNN

198/ 80–20 split,

[35] MUMA Biological risk factor KNN 84.1% accuracy
400 mixed user

198/ Raw time 80–20 split,

[36] MUMA LSTM 92% accuracy
400 series window mixed user

SmartFall, 66–33 split, 85% accuracy,

221/
[20] Notch, Time domain SVM mixed user, 93% accuracy,
30000
Farseeing LOO CV 99% accuracy

SmartFall, 221/ Raw time 80–20 split,

[32] CNN 99.7% accuracy
Notch 30000 series window mixed user

1798/ 10 Fold CV,

[37] SisFall Time domain SVM 99.75% accuracy
2707 mixed user

Raw time
[38] SisFall Not specified LSTM Not specified 97.16% accuracy
series window

1575 Falls/ 5 Fold CV,

[39] SisFall Time domain SVM with RBF 98% accuracy
not specified mixed user

1780 Falls/ Raw time CNN and 2 Fold CV,

[33] SisFall 99.94% accuracy
76926 ADL series window various other separated user

2.4. 1D Convolutional Neural Networks

Convolutional neural networks (CNN) were first introduced by Krizhevsky et al. [40] and gained
attention by winning the ImageNet challenge [41] with a large margin in 2012. Since then, new CNN
Architectures, e.g., ResNet [42] were introduced and enhanced the performance of CNN significantly.
CNN’s were mostly used for the image domain. In 2016, Wang et al. [43] introduced a fully
convolutional neural network (FCN) architecture to classify time series data. They validated their
approach on 44 datasets from the UCR/UEA archive. CNN for time series data processing achieved
state of the art performances in various domains e.g., ECG Classification, Sound Classification,
and Natural Language Processing. The name giving convolution can be seen as applying and sliding
a one-dimensional filter over the time series. Unlike images, the filters cover only one dimension
(time) opposed to two dimensions (width and height). A convolutional neural network consists of
various filters, ranging from moving average filters to more complex filters. Those filters are learned
through the backpropagation algorithm. By passing a univariate time series through a convolutional
neural network, multiple convolutions with different filters are applied. This may be seen as using
a filter bank to extract useful features, removing outliers or general filtering. To surpass the linear
nature of a convolution operation, a nonlinear function is applied after a convolution to introduce
nonlinearity and therefore ensure a nonlinear transformation of the time series data. An illustration of a
Technologies 2020, 8, 72 5 of 17

one-dimensional convolutional neural network is denoted in Figure 1. Mathematically, a convolutional

neural network apply the following equation for each timestamp t of the time series data X:

Ct = f (W ∗ Xt− k :t+ k + b)
2 2

were W, k, t, b equals weights of the kernel, length of the kernel, timestamp, and bias, respectively.
After applying n convolutions on the input accelerometer data with length l, we settle with n channels,
where each channel represents a new filtered time series. These n channels with shape n × l are then
convolved with m different filters with shape n × m × k, where each mi−th filter is slid across all n
channels resulting in m additional time series, where each mi channel is a sum of the convolutions of
mi−th filter across all n channels.

Figure 1. Illustration of a one-dimensional convolutional neural network.

2.5. Recurrent Neural Networks

Recurrent Neural Networks (RNN) are designed to work with time series data. Their recurrent
mechanisms ensure that features in the time series data are learned jointly over time. To train a recurrent
neural network, the back-propagation through time algorithm is used. Recurrent neural networks
emerged in 1986 [44] and underlie various advances. One major drawback of traditional recurrent
neural network is the vanishing or exploding gradient problem, which occur when the gradient
of each cell is multiplied over time and gets larger or smaller with each multiplication. Training a
recurrent neural network was a hard task to complete. Long Short-Term Memory (LSTM) [45] neural
networks eliminated the problem of vanishing and exploding gradients by introducing forget gates.
This mechanism enabled the effective training of recurrent neural network models.

2.5.1. Long Short-Term Memory

One major difference to traditional RNN models is essentially what a Long Short-Term Memory
(LSTM) learns. It can learn to keep only relevant information to make predictions and forget non
relevant data. This renders the LSTM easier to train and perform better on classification or regression
tasks. This process is made through a sigmoid-like function, which outputs a value between zero
(discard) and one (keep).

3. Experimental Setup

3.1. Data Preparation

In order to combine datasets containing wrist-worn accelerometer data, a series of steps
is employed to utilize them for training. Due to different sampling rates and sensors, several datasets
need special handling. Some datasets include a label for each timestamp, while others provide a sample
Technologies 2020, 8, 72 6 of 17

as a text file with the label being part of the files name. For each recording (containing either falls
or activities of daily living) in a data set, we segment the recording into 10 s of non overlapping
windows and down-or upsample each window to 50 Hz. The range and quantization of the raw data
remain unchanged. The variety in coupling between human body and sensor, precisely the attachment
of the sensor at the wrist, and also different sensor weights are not considered. Each fall type is labeled
as fall and every other activity is labeled as not fall.

3.1.1. Down and Upsampling Technique

In order to perform robust down-and upsampling, we use a polyphase filtering approach
provided by the Python SciPy 1.2.3 package [46]. The acceleration signal is upsampled by a factor up,
before a zero-phase low-pass FIR filter is applied, and then again downsampled by the factor down.
The resulting sample rate is up/down times the original sample rate. We pad the signal boundaries
by fitting a line at the start and end of the signal to avoid overshooting at the boundaries of the signal.
We are aware that a down- and upsampling of data are somehow affecting the data quality, but this
limitation is necessary for a comparable dataset.

3.1.2. SmartFall, Smartwatch, and Notch

For the SmartFall, Smartwatch, and Notch datasets, the following steps are applied:

1. Cluster timestamps containing a fall to create separated 10 s segments.

2. Each segment is upsampled from 310 (31 Hz) to 500 timestamps (50 Hz)

After applying this strategy, SmartFall, Smartwatch, and Notch consist of solely fall samples.
We do not use samples where a fall sample is mixed with an activity of daily living labeled as non
fall. This would result in an ambiguity, as we already use samples containing fall and activities of
daily living labeled as fall. To enrich the Notch, Smartwatch, and SmartFall datasets with additional
activities of daily living, we added 500 random 10 s segments from the RealWorld Human Activity
Recognition [47] dataset to the SmartFall, Smartwatch, and Notch datasets.

3.1.3. MUMA, UP Fall, and Sim Fall

The MUMA dataset consists of separated files containing either fall or activities of daily living,
where each file consists of 300 timestamps. We cropped a 200 timestamp window (10 s) out of the time
series and upsampled it to 500 timestamps. To harmonize the UP Fall dataset, we applied a slightly
different strategy, as this dataset does not offer a file for each fall or activity of daily living. A peak
detection algorithm provided by the Python SciPy 1.2.3 package [46] is used to find the segments
containing a fall, where we set the prominence to a 95% quantile of the Signal Magnitude Vector values.
We crop a 180 timestamp window, centered around the detected peaks, and upsampled this window
to 500 timestamps. The Sim Fall dataset consists of separated files containing either data of fall or
activities of daily living. We crop a 250 timestamp window, centered around the center corresponding
time series.

3.2. Harmonized Dataset

The resulting harmonized dataset consists of 1716 Falls and 3567 activities of daily living,
where each sample consists of 500 timestamps associated with x, y, and z-axis acceleration values.
We settle with a total tensor shape of 5283 × 500 × 3 for our experiments. Table 4 shows the number of
falls and activities of daily living for each dataset. Note that the number of activities of daily living
is significantly larger than the number of falls. To compensate such unbalanced dataset, we induce
a weight to the cross entropy loss function calculated by the ratio of falls and activities of daily
living. Note that we did not integrate the TST Fall Detection dataset, as it is not available anymore
and it is not clear which sensor is used on the wrist and which sensor is used on the waist.
Technologies 2020, 8, 72 7 of 17

Table 4. Harmonized dataset.

Dataset Falls ADL Subjects Age of Subjects

MUMA 189 428 17 18–55
Notch 198 500 7 20–35
Sim Fall 286 258 17 19–27
Smartwatch 271 500 7 21–55
SmartFall 527 500 14 21–60
UP Fall 245 1381 17 18–24
Total 1716 3567 79 18–60

3.3. Problem Formulation

The fall detection or fall classification problem may be formulated as a problem of time series
classification. A fall may represent a univariate time series X = [ x1 , x2 , . . . , xt ] with an ordered
set of real values or a multivariate time series M = [ X 1 , X 2 , . . . , X n ], where M consists of different
univariate time series X. A dataset may be formed by pairing the time series data X with a label Y
to a tuple ( X i , Y i ). This label is a numerical representation of a class label either fall with Y i = 1 or
non fall with Y i = 0. The task is to train a classifier on a dataset consisting of multiple time series
in order to map the space of possible inputs to a probability distribution over the class variable values,
referred to as labels.

3.4. Preprocessing
Contrary to other approaches e.g., [32], we do not think that a scaling or standardization
preprocessing technique should be performed on the input data, as the magnitude information
is a crucial aspect to distinguish between falls and activities of daily living. To evaluate the impact
of preprocessing on the performance, we apply a min-max scaling of the training and testing dataset.

Signal Magnitude Vector (SMV)

To reduce the computational load, we solely use the acceleration data, precisely the Signal
Magnitude Vector. The Signal Magnitude Vector maps the three-dimensional acceleration vector
X ∈ R3 to a one-dimensional acceleration vector with no orientation information. To compute
the Signal Magnitude Vector, the following equation is applied for each time series X:
q
SMV = a2x + a2y + a2z

The Signal Magnitude Vector transformation greatly reduces the input size for neural networks
and thus reducing the required memory and computational load of an IoT device. Note that running
deep learning models on IoT devices requires a trade-off between classification performance
and computational complexity.

3.5. Data Augmentation

Data Augmentation can be viewed as an injection of prior knowledge about the invariant
properties of the data against certain transformations. Augmented data can cover unexplored input
space, prevent overfitting, and improve the generalization ability of a deep learning model [48].
Um et al. pointed out that time series augmentation significantly improves the classification accuracy
of neural network architectures from 77.54% to 86.88 % in the domain of Parkinson’s Disease
classification in Alzheimer patients [49]. In the domain of fall detection, augmentation showed
already an enormous performance boost in other works [32]. Before augmentation, one requires
knowledge about the semantic characteristics of the input data. Scaling of the input data may induce
bias, as falls typically have a defined range of acceleration values. To evaluate the effect of data
augmentation, we apply a combination of the following transformations dynamically during training:
Technologies 2020, 8, 72 8 of 17

• Shifting with a probability of p = 0.75 and random shift value of [−150 . . . 150] in samples along
the time axis (only used during signal magnitude vector training)
• Rotation with a probability of p = 0.75 around x, y, and z-acceleration axes with a random angle
between [−180◦ . . . 180◦ ] (only used during three-axis training)

We suspect that shifting in the time dimension does not contribute to the performance of a
one-dimensional CNN, as the learned filters are translation-invariant by definition. However, by shifting
(rolling) the vector in the time dimension, elements that roll beyond the last position of the vector
are re-introduced at the first position. This induces some additional variance to data and may affect
the performance of the neural networks.

3.6. Model Architectures

To compare our IoT neural network to other neural network architectures, we choose a number
of different architecture types. We utilized the PyTorch 1.2.0 Framework [50] for training and testing
our approach. All models were trained 250 epochs each with the Adam Optimizer [51] with initial
learning rate of 0.001, batch size of 32, β 1 = 0.9, β 2 = 0.999 and no weight decay using a weighted
Cross Entropy loss function. As our convolutional neural networks use ReLU (Rectifier Linear Unit)
activation functions, all convolutional layers were initialized via He initialization [52]. The ReLU is
a nonlinear activation function which is defined as σ ( x ) = max (0, x).

3.6.1. Classic 1D CNN

CNN-3B3Conv [32] consists of three-layer blocks. The first block consists of three convolutional
layers and one maxpooling layer. Each of the convolutional layer consists of 64 kernels with a size
of 4 and a maxpooling size of 2. The second block also consists of three convolutional layers and one
maxpooling layer, this time with a kernel size of 3 while the maxpooling size remains unchanged.
The third block consists of three fully-connected layers with 64 neurons, 32 neurons, and two neurons.
To deal with the different input size (500 opposed to 32), we changed the pooling size of the maxpooling
parameter to 15.

3.6.2. 1D ResNet
We use a small-scale 1D ResNet model consisting of two basic residual connection blocks. We settle
with a convolutional layer followed by two blocks with two convolutional layers and a shortcut
connection each block. The kernel size is set to 8 for the first convolutional layer and is changed to four
for all consecutive convolutional layers. The classification head consists of a fully connected layer
mapping 64 neurons to two neurons.

3.6.3. LSTM
The LSTM Classifier consists of a single unidirectional LSTM layer with 128 hidden neurons
followed by a dropout layer with a dropout probability of p = 0.25. A fully connected layer is used to
map the 128 hidden neurons to two neurons for classification.

3.7. Proposed Model Architecture

Our baseline CNN for fall classification includes several changes compared to Santos et al. [32].
As Santos et al. pointed out in their evaluation, a deeper network does not necessary increase
the performance of classification. Our approach to fall detection consists of greatly reducing the number
of layers and filters of the convolutional neural network. Furthermore, we replaced the computational
expensive max-pooling layer with efficient strided convolutional layer [53]. By removing the flattening
layer right before the fully connected layer and replacing them with a global average pooling (GAP)
layer followed by a 1 × 1 convolutional layer, we further reduced the number of trainable parameter
significantly. To effectively deploy our model to IoT devices, we additionally use quantization
Technologies 2020, 8, 72 9 of 17

which reduce the footprint of the model drastically. After employing an extensive gridsearch, we settle
with the following structure:

C (8) − ReLU − C (16) − ReLU − C (32) − ReLU − GAP − C (2)

Each convolutional layer C ( N ), with N channels, uses a stride of 4 and a kernel size of 8 with no
padding, except for the last convolutional layer which uses stride and kernel size of 1 with no padding.

3.7.1. Maxpooling vs. Strided Convolutions

Strided Convolutions reduce, similar to maxpooling, the size of the resulting feature map and thus
reduce the computational complexity of the processed data. We replaced the maxpooling layers with
convolutional layers with stride 4 to eliminate computational expensive calculation of the maximum
value and additionally reduce the number of dot products by a factor of 4. A strided convolution
calculates the dot product for every n − th timestamp by skipping n timestamps with each calculation
resulting in a downsampling by a factor of n.

3.7.2. Global Average Pooling Layer

Global Average Pooling (GAP) allows the Convolutional Neural Network to process time series
with different lengths and reduces the number of learned parameters significantly [54]. Furthermore,
a GAP layer enables us to use Class Activation Maps for visualization. Global Average Pooling
calculates the mean for each channel (activation map) resulting in a vector of averaged activation
values of each channel.

3.7.3. Learnable Parameters

The amount of learnable parameters affect the training time, computational complexity, and
footprint on the device. By replacing the flattening layer with GAP and Maxpooling layer with strided
convolutions, we reduced the number of learnable parameters to around 5000 opposed to 72.512 in [32]
and further reduced the required memory for storing the activation maps (channels) significantly.
Our model is approx. 14 times smaller than [32] and can be easily used on small scale IoT devices as it
only uses approx. 20 KB of RAM for storing the weights and approx. 10 KB for storing the activation
values for each channel.

3.7.4. Quantization
Quantization reduces the footprint of a model greatly, by converting the weights of the neural network
from four byte floats to one byte unsigned integer. Wu et al. [55] showed that a quantization occurs with
a minor loss in accuracy while reducing the computation time greatly. We use a Quantization method
provided by the PyTorch 1.4.0 Framework, in particular the min and max values to compute the necessary
quantization parameters. We do not evaluate different quantization strategies to conserve space.

4. Evaluation

4.1. Evaluation Method

In order to employ a robust evaluation process, our approach consists of applying Leave One Out
(LOO) cross validation on whole datasets, e.g., using the MUMA Dataset for testing and the remaining
five wrist-worn datasets for training. We reset the weights and the optimizer parameters with each fold.
This procedure is repeated five times until each dataset has been used as a testing dataset. To discard
potential weight initialization bias, we repeated the LOO cross validation five times and averaged
the results.
Technologies 2020, 8, 72 10 of 17

4.1.1. Evaluation Metric

We assess the classification performance of our deep learning model by using the weighted F1
score, precision, and recall. Note that, in the binary classification task, recall of the positive class (fall)
is denoted as sensitivity and recall of the negative class (not fall) as specificity.

TP TP precision · recall
precision = recall = F1 = 2 ·
TP + FP TP + FN precision + recall

4.2. Effect of Augmentation

In line with other work, we observed performance improvements while using data augmentation
techniques for all tested models. The results achieved with data augmentation are denoted in Tables 5
and 6. While we observe that our three-axis model increased in terms of performance, our single-axis
IoT-CNN achieved only little or negligible improvement. Larger models, e.g., CNN-3B3Conv,
improved in performance while using dynamic data augmentation during three-axis and single-axis
training training.

4.3. Learned Filters

To visualize the frequency response of the learned filters in the first convolutional layer, we use
the freqz method, provided by the Python package scipy [46]. A lot of filters resemble low pass,
bandstop, high pass, and further combination of low and highpass filters. The learned kernel weights
with their respective frequency response are denoted in Figures 2 and 3. Note that most learned filters
activate on strong edge shaped structures.
Frequency Response from learned Filter 0 Frequency Response from learned Filter 1 Frequency Response from learned Filter 2 Frequency Response from learned Filter 3 Frequency Response from learned Filter 4 Frequency Response from learned Filter 5 Frequency Response from learned Filter 6 Frequency Response from learned Filter 7
6 5
0 0 5.0
0 0
0 4 0
5 2.5
5 5
10 0.0 10 2 5
5
10
Amplitude [dB]

Amplitude [dB]

Amplitude [dB]

Amplitude [dB]

Amplitude [dB]

Amplitude [dB]

Amplitude [dB]

Amplitude [dB]
15 2.5
10 0
20 5.0 20 15 10
10
2
25 15 7.5 20
30 4 15
30 10.0 15 25
20 6 30
35 12.5
40 20
20
0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25
Frequency [Hz] Frequency [Hz] Frequency [Hz] Frequency [Hz] Frequency [Hz] Frequency [Hz] Frequency [Hz] Frequency [Hz]

(a) (b) (c) (d) (e) (f) (g) (h)

Figure 2. Frequency responses of the first convolutional layer.

Weights from learned Filter 0 Weights from learned Filter 1 Weights from learned Filter 2 Weights from learned Filter 3 Weights from learned Filter 4 Weights from learned Filter 5 Weights from learned Filter 6 Weights from learned Filter 7
0.5
0.4 0.5 0.4
0.4 0.4
0.5 0.4
0.3 0.4 0.3
0.3 0.3 0.4
0.4 0.2 0.2
0.2 0.2 0.2 0.3
0.3
0.3 0.1 0.1
0.1 0.1 0.2 0.2
0.2 0.0 0.0 0.0
0.0 0.0
0.1 0.1 0.1
0.1 0.1 0.1
0.1 0.2
0.0 0.0 0.2
0.2 0.0 0.2
0.2
0.3
0.3 0.3 0.1 0.1
0.1 0.4
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7

(a) (b) (c) (d) (e) (f) (g) (h)

Figure 3. Learned kernel weights of the first convolutional layer.

Technologies 2020, 8, 72 11 of 17

GT: 1 PRED: 1 GT: 1 PRED: 1 GT: 1 PRED: 1

7 25
1.75 10
20
0.12
1.50 6 20
8 15
0.10
1.25 5 15
1.00 0.08 6 10 4
10
0.75 0.06 5 3
4
5
0.50 2
0.04
2 0 0
0.25 1
0.02
5
0 100 200 300 400 500 0 100 200 300 400 500 0 100 200 300 400 500

(a) Notch (b) MUMA (c) Sim Fall

GT: 1 PRED: 1 GT: 1 PRED: 1 GT: 1 PRED: 1

20
20
6 6 7
15 15
5 6 15
5
10 5
4 10
4 10
4
3 3 5
5 3 5
2
2 0 2
0 0
1
1 1
5
5
0 100 200 300 400 500 0 100 200 300 400 500 0 100 200 300 400 500

(d) SmartWatch (e) SmartFall (f) UP Fall

Figure 4. Class Activation Map examples for fall data sampled from each dataset.

4.4. Class Activation Maps

Class Activation Maps (CAM) for time series data, introduced by Zhou et al. in [56], indicate
which regions in a time series contribute to the decision-making process of a neural network. A Class
Activation Map (CAM) for a class c may be computed by the following equation:

CAMc (t) = ∑ wmc Am (t)

m

where Am (t) is the univariate time series for the variable m ∈ [1, M], which is in fact the result
c is the weight between the m − th filter (last convolutional layer)
of applying the m − th filter and wm
and the output neuron of class label c [54]. Examples of class activation maps with a sample for
each dataset are depicted in Figure 4. In most cases, the impact region (region with the highest
acceleration magnitude) and free fall phase (region right before the impact) are the most contributing
parts to the decision-making of our convolutional neural network.

(ResNet, AUC) (Santos, AUC) (LSTM, AUC) (ResNet, AUC) (Santos, AUC) (LSTM, AUC)
(FallNet, AUC) (FallNet, AUC)

1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
Notch MUMA Sim Fall SmartWatch SmartFall Up Fall Mean Notch MUMA Sim Fall SmartWatch SmartFall Up Fall Mean

(a) AUC 1-Axes with Data Augmentation (b) AUC 3-Axes with Data Augmentation

Figure 5. AUC Results with data augmentation for single-axis (left) and three-axis (right). Bar charts
showing the AUC performance for each dataset and network.
Technologies 2020, 8, 72 12 of 17

4.5. Results
Referring to Tables 5 and 6, we demonstrated that our proposed algorithm performs very accurate
by an F1 = 0.96 without, and F1 = 0.97 with data augmentation (both on SMV). The results in Figure 5
suggest that a larger number of parameters does not lead to increased performance in terms of AUC.
Furthermore, the three-axis models do not perform better than their single-axis equivalent. The LSTM
neural network especially shows increasing performance when using the Signal Magnitude Vector as
input. In particular, the three-axis version of our IoT-CNN does perform poorly on the MUMA dataset
and SimFall dataset if no augmentation is applied. We suspect that this may be due to the relative
small number of parameters compared to other models. All CNNs perform better on the SmartWatch,
SmartFall, Notch, and UP Fall datasets. This indicates that certain datasets are easier to handle than
other and may contain less variety. Note that the SmartWatch, SmartFall, and Notch dataset was
published by the same researchers. We further augmented the SmartFall, Smartwatch, and Notch
dataset with samples of a different dataset. This is done due to the sequence length of a fall in these
datasets, as they only contain fall events shorter than 2 s. Regarding the LSTM-based neural network,
the single-axis version shows a comparable performance with respect to their CNN counterparts.
Quantization, on the other hand, achieved, as expected, worse results. In line with other work,
quantization reduces the classification performance (see Table 7). While the performance on most
datasets remains comparable to the results without quantization, the performance on the Notch dataset
decreases by a large amount. This may be due to the gravitational offset from the used sensor.

Table 5. Results (F1 Score) without Data Augmentation.

Our LSTM Santos ResNet

Dataset SMV 3 Axis SMV 3 Axis SMV 3 Axis SMV 3 Axis
µ σ µ σ µ σ µ σ µ σ µ σ µ σ µ σ
Notch 0.98 0.01 0.97 0.03 1.00 0.00 0.95 0.01 0.92 0.04 0.94 0.01 0.77 0.14 0.91 0.05
MUMA 0.92 0.00 0.80 0.04 0.92 0.01 0.75 0.04 0.91 0.01 0.91 0.01 0.90 0.01 0.89 0.02
Sim Fall 0.89 0.00 0.83 0.04 0.88 0.02 0.80 0.12 0.88 0.01 0.89 0.00 0.90 0.01 0.89 0.01
SmartWatch 0.99 0.00 1.00 0.00 1.00 0.00 0.95 0.04 1.00 0.00 1.00 0.00 1.00 0.00 1.00 0.00
SmartFall 0.99 0.00 1.00 0.00 0.99 0.00 0.92 0.02 1.00 0.00 1.00 0.00 0.99 0.00 1.00 0.00
UP Fall 0.98 0.00 0.90 0.02 0.98 0.00 0.78 0.07 0.99 0.00 0.97 0.00 0.99 0.00 0.96 0.02
Average 0.96 0.00 0.91 0.02 0.96 0.01 0.86 0.05 0.95 0.01 0.95 0.00 0.93 0.03 0.94 0.02

Table 6. Results (F1 Score) with Data Augmentation.

Our LSTM Santos ResNet

Dataset SMV 3 Axis SMV 3 Axis SMV 3 Axis SMV 3 Axis
µ σ µ σ µ σ µ σ µ σ µ σ µ σ µ σ
Notch 0.97 0.01 0.95 0.03 1.00 0.00 0.93 0.04 1.00 0.00 0.96 0.01 0.97 0.03 0.93 0.01
MUMA 0.93 0.01 0.94 0.01 0.91 0.01 0.84 0.06 0.92 0.00 0.92 0.00 0.92 0.01 0.94 0.01
Sim Fall 0.93 0.00 0.90 0.01 0.93 0.01 0.81 0.10 0.89 0.00 0.90 0.00 0.94 0.00 0.90 0.01
SmartWatch 0.99 0.00 1.00 0.00 0.99 0.00 0.95 0.03 1.00 0.00 1.00 0.00 1.00 0.00 1.00 0.00
SmartFall 0.99 0.00 1.00 0.00 0.98 0.00 0.92 0.03 1.00 0.00 1.00 0.00 1.00 0.00 1.00 0.00
UP Fall 0.98 0.00 0.99 0.00 0.97 0.00 0.91 0.05 0.98 0.00 0.99 0.00 0.98 0.00 0.99 0.00
Average 0.97 0.00 0.96 0.01 0.96 0.01 0.89 0.05 0.97 0.00 0.96 0.00 0.97 0.01 0.96 0.01
Technologies 2020, 8, 72 13 of 17

Table 7. Results (Precision, Recall, and Weighted F1 Score) using our CNN with Data Augmentation
and Weight Quantization.

Our
SMV
Dataset
Precision Recall F1
µ σ µ σ µ σ
Notch 0.854 0.092 0.853 0.078 0.826 0.110
MUMA 0.926 0.017 0.910 0.024 0.912 0.023
Sim Fall 0.908 0.016 0.902 0.023 0.902 0.023
SmartWatch 0.982 0.004 0.982 0.004 0.982 0.004
SmartFall 0.976 0.009 0.975 0.010 0.975 0.010
UP Fall 0.952 0.013 0.940 0.035 0.942 0.029
Average 0.933 0.025 0.927 0.029 0.923 0.033

5. Discussion
Historically, fall detection algorithms were discrete and engineers developed a mobile system
especially for this purpose. Today, neural networks can be exploited on embedded systems, so a flexible
structure can be used. Because of the connectivity of IoT devices, a broad database is or will be available
in future. Therefore, the realization of a reliable fall detection becomes available. The performance
consideration of fall detection is ambiguous as far as varying datasets are used. Therefore, we identified
available datasets and evaluated the capabilities of a small-scale neural network. Large networks that can
run on servers or high end machines might outperform small-scale neural networks, but these can be
integrated on edge or small devices and are more relevant for the real life scenarios, as the computation
is done on the device. This further ensures (a) that sensible data are processed on the device and
no network connection is needed, besides the alarming mechanism, and (b) that privacy is protected.
Regarding the existing fall datasets, we assume that resampling, range, resolution, and sensor type
(internal filtering) only have a minor effect. However, this has to be confirmed by further research.

6. Outlook and Application

Typical deep neural network models require high computational power for training a large scale
neural network. However, the inference is much less energy consuming. As we could show in our
research, even small models provide a reasonable recognition rate of 97%. The representation of weights
can be shrunk from 32-bit floats to unsigned 8-bit integers without a remarkable loss of accuracy and
reduction in the memory footprint on the device. By levering compact architecture with quantized
weights, we are able to use our efficient neural network on embedded devices. This research forms
the center of the lightweight wrist-based fall detection device, called UMA, designed by the company
Next Step Dynamics, which has a long battery life as the fall detection is processed in a dual recognition
pipe—an acceleration sensor monitors all movements and a small size trigger algorithm works as
a gatekeeper and detects very roughly possible falls. Due to the dual recognition pipe, a possible
requirement of a fast processing speed of the neural net is not necessary. Furthermore, the trigger
algorithm ensures a continuous processing and provides easy to handle segmented data to the second
inference step. The device (based on a Nordic nRF9160) is shown in Figure 6. We implemented our
proposed neural network on the UMA fall detection device and validated its capabilities in a real-life
scenario in future work.
Technologies 2020, 8, 72 14 of 17

Figure 6. UMA Device with an integrated neural net for fall detection.

Elderly persons are usually very calm; even a normal adult is performing moderate to vigorous
leisure time physical activity only less than one hour a day [57]. The simple gatekeeper algorithm
clears inactive periods, so the deep learning algorithm is used only for active periods.

7. Conclusions
In this paper, we presented 19 datasets with fall raw data, assessed at the wrist (7), waist (10),
and/or other positions (2) with accelerometers. We illustrated that our optimized neural network could
be applied on an embedded system like IoT devices, smart watches, or activity trackers. In former
research, the consideration of the accuracy of fall detection was performed within a focus group
and its dataset. Because of the identification of multiple datasets, a broader evaluation was feasible.
We could show that neural networks are performing well on our harmonized dataset. Furthermore,
the increasing calculation power of mobile devices enables the usage of deep learning algorithm for
fall detection on wrist based embedded systems. We demonstrated that small scale convolutional
neural networks achieve a reasonable accuracy of 97% on our harmonized fall detection data set.
While applying quantization, our neural network performs less accurately, which may be addressed
in future work. We suspect that calibration of the neural network based on the activity of the user may
enhance the performance significantly by lowering the false positive rate. For future work, we see that
the sensitivity and specificity of fall detection is highly relevant in the everyday usage. A waterproof,
wrist based sensing device can be worn 24/7 and should indicate no false detection. This is a high
demand and requires the knowledge about the general condition about the user. Very active people
are moving differently compared to passive and calm people. This requirement of a low false positive
rate can be achieved by individualized algorithms or calibration. We assume that even energy efficient
mobile wrist devices allow a reliable fall detection system to assist the elderly in everyday life.

Author Contributions: Conceptualization, D.K. and G.B.; methodology, G.B. and K.S.; software, D.K.;
data curation, D.K.; writing—original draft preparation, D.K., K.S. and G.B.; writing—review and editing,
D.K., K.S. and G.B.; supervision, G.B. All authors have read and agreed to the published version of the manuscript.
Funding: The research was funded by Next Step Dynamics AB..
Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Stevens, J.A.; Corso, P.S.; Finkelstein, E.A.; Miller, T.R. The costs of fatal and non-fatal falls among older
adults. Inj. Prev. 2006, 12, 290–295. [CrossRef]
2. Chaudhuri, S.; Thompson, H.; Demiris, G. Fall detection devices and their use with older adults:
A systematic review. J. Geriatr. Phys. Ther. 2014, 37, 178–196. [CrossRef]
Technologies 2020, 8, 72 15 of 17

3. Boyle, J.; Karunanithi, M. Simulated fall detection via accelerometers. In Proceedings of the 2008 30th
Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Vancouver, BC,
Canada, 20–25 August 2008; pp. 1274–1277.
4. Nunez-Marcos, A.; Azkune, G.; Arganda-Carreras, I. Vision-based fall detection with convolutional neural
networks. Wirel. Commun. Mob. Comput. 2017, 2017, 9474806. [CrossRef]
5. Alwan, M.; Rajendran, P.J.; Kell, S.; Mack, D.; Dalal, S.; Wolfe, M.; Felder, R. A smart and passive
floor-vibration based fall detector for elderly. In Proceedings of the 2006 2nd International Conference on
Information & Communication Technologies, Damascus, Syria, 24–28 April 2006; Volume 1, pp. 1003–1007.
6. Daher, M.; Diab, A.; El Najjar, M.E.B.; Khalil, M.A.; Charpillet, F. Elder tracking and fall detection system
using smart tiles. IEEE Sens. J. 2016, 17, 469–479. [CrossRef]
7. Casilari, E.; Santoyo-Ramón, J.A.; Cano-García, J.M. UMAFall: A Multisensor Dataset for the Research on
Automatic Fall Detection. Procedia Comput. Sci. 2017, 110, 32–39. [CrossRef]
8. Frank, K.; Vera Nadales, M.J.; Robertson, P.; Pfeifer, T. Bayesian recognition of motion related activities with
inertial sensors. In Proceedings of the 12th ACM International Conference Adjunct Papers on Ubiquitous
Computing—Adjunct, Copenhagen, Denmark, 26–29 September 2010; Bardram, J.E., Ed.; ACM: New York, NY,
USA, 2010; p. 445. [CrossRef]
9. Vavoulas, G.; Pediaditis, M.; Spanakis, E.G.; Tsiknakis, M. The MobiFall dataset: An initial evaluation of
fall detection algorithms using smartphones. In Proceedings of the IEEE 13th International Conference on
Bioinformatics and Bioengineering (BIBE), Chania, Greece, 10–13 November 2013; pp. 1–4. [CrossRef]
10. Gasparrini, S.; Cippitelli, E.; Gambi, E.; Spinsante, S.; Wåhslén, J.; Orhan, I.; Lindh, T. Proposal and
Experimental Evaluation of Fall Detection Solution Based on Wearable and Depth Data Fusion. In ICT
Innovations 2015; Advances in Intelligent Systems and Computing; Loshkovska, S.; Koceski, S., Eds.;
Springer: Cham, Switzerland, 2016; Volume 399, pp. 99–108. [CrossRef]
11. Medrano, C.; Igual, R.; Plaza, I.; Castro, M. Detecting falls as novelties in acceleration patterns acquired
with smartphones. PLoS ONE 2014, 9, e94811. [CrossRef]
12. Kwolek, B.; Kepski, M. Human fall detection on embedded platform using depth maps and wireless
accelerometer. Comput. Methods Programs Biomed. 2014, 117, 489–501. [CrossRef]
13. Özdemir, A.T.; Barshan, B. Detecting falls with wearable sensors using machine learning techniques. Sensors
2014, 14, 10691–10708. [CrossRef]
14. Ojetola, O.; Gaura, E.; Brusey, J. Data set for fall events and daily activities from inertial sensors.
In Proceedings of the 6th ACM Multimedia Systems Conference, Portland, OR, USA, 18–20 March 2015;
Ooi, W.T., Feng, W., Liu, F., Eds.; ACM: New York, NY, USA, 2015; pp. 243–248. [CrossRef]
15. Vilarinho, T.; Farshchian, B.; Bajer, D.G.; Dahl, O.H.; Egge, I.; Hegdal, S.S.; Lones, A.; Slettevold, J.N.;
Weggersen, S.M. A Combined Smartphone and Smartwatch Fall Detection System. In Proceedings of
the CIT/IUCC/DASC/PICom 2015, Liverpool, UK, 26–28 October 2015; pp. 1443–1448. [CrossRef]
16. Wertner, A.; Czech, P.; Pammer-Schindler, V. An Open Labelled Dataset for Mobile Phone Sensing Based
Fall Detection. In Proceedings of the 12th EAI International Conference on Mobile and Ubiquitous Systems:
Computing, Networking and Services, Coimbra, Portugal, 22–24 July 2015; Zhang, P., Silva, J.S., Lane, N.,
Boavida, F., Rodrigues, A., Eds. MobiQuitous: Coimbra, Portugal, 2015. [CrossRef]
17. Aziz, O.; Musngi, M.; Park, E.J.; Mori, G.; Robinovitch, S.N. A comparison of accuracy of fall detection
algorithms (threshold-based vs. machine learning) using waist-mounted tri-axial accelerometer signals
from a comprehensive set of falls and non-fall trials. Med. Biol. Eng. Comput. 2017, 55, 45–55. [CrossRef]
18. Sucerquia, A.; López, J.D.; Vargas-Bonilla, J.F. SisFall: A Fall and Movement Dataset. Sensors 2017, 17, 198,
[CrossRef]
19. Micucci, D.; Mobilio, M.; Napoletano, P. UniMiB SHAR: A Dataset for Human Activity Recognition Using
Acceleration Data from Smartphones. Appl. Sci. 2017, 7, 1101. [CrossRef]
20. Mauldin, T.R.; Canby, M.E.; Metsis, V.; Ngu, A.H.H.; Rivera, C.C. SmartFall: A Smartwatch-Based Fall
Detection System Using Deep Learning. Sensors 2018, 18, 3363. [CrossRef]
21. Chan, H.L. CGU-BES Dataset for Fall and Activity of Daily Life. Figshare 2018. [CrossRef]
22. Martínez-Villaseñor, L.; Ponce, H.; Brieva, J.; Moya-Albor, E.; Núñez-Martínez, J.; Peñafort-Asturiano, C.
UP-Fall Detection Dataset: A Multimodal Approach. Sensors 2019, 19, 1988, [CrossRef]
23. Cotechini, V.; Belli, A.; Palma, L.; Morettini, M.; Burattini, L.; Pierleoni, P. A dataset for the development and
optimization of fall detection algorithms based on wearable sensors. Data Brief 2019, 23, 103839. [CrossRef]
Technologies 2020, 8, 72 16 of 17

24. Mauldin, T.; Ngu, A.H.; Metsis, V.; Canby, M.E.; Tesic, J. Experimentation and analysis of ensemble deep
learning in iot applications. Open J. Internet Things 2019, 5, 133–149.
25. Pannurat, N.; Thiemjarus, S.; Nantajeewarawat, E. Automatic Fall Monitoring: A Review. Sensors 2014,
14, 12900–12936. [CrossRef]
26. Krupitzer, C.; Sztyler, T.; Edinger, J.; Breitbach, M.; Stuckenschmidt, H.; Becker, C. Hips Do Lie!
A Position-Aware Mobile Fall Detection System. In Proceedings of the 2018 IEEE International Conference
on Pervasive Computing and Communications (PerCom); Communications, Athens, Greece, 19–23 March
2018; pp. 1–10. [CrossRef]
27. Liao, M.; Guo, Y.; Qin, Y.; Wang, Y. The application of EMD in activity recognition based on a single triaxial
accelerometer. Bio-Med. Mater. Eng. 2015, 26 ( Suppl. 1), S1533–S1539. [CrossRef]
28. Bourke, A.K.; O’Brien, J.V.; Lyons, G.M. Evaluation of a threshold-based tri-axial accelerometer fall detection
algorithm. Gait Posture 2007, 26, 194–199. [CrossRef]
29. Kangas, M.; Konttila, A.; Winblad, I.; Jämsä, T. Determination of simple thresholds for accelerometry-based
parameters for fall detection. In Proceedings of the Annual International Conference of the IEEE Engineering
in Medicine and Biology Society, Lyon, France, 22–26 August 2007; pp. 1367–1370. [CrossRef]
30. Salomon, R.; Lueder, M.; Bieber, G. iFall—A New Embedded System for the Detection of Unexpected
Falls. In Proceedings of the Eighth Annual IEEE International Conference on Pervasive Computing
and Communications Workshops (PerCom Workshops), Mannheim, Germany, 29 March–2 April 2010.
[CrossRef]
31. Vallabh, P.; Malekian, R.; Ye, N.; Bogatinoska, D.C. Fall detection using machine learning algorithms.
In Proceedings of the 2016 24th International Conference on Software, Telecommunications and Computer
Networks (SoftCOM), Split, Croatia, 22–24 September 2016; pp. 1–9. [CrossRef]
32. Santos, G.L.; Endo, P.T.; Monteiro, K.H.C.; Rocha, E.S.; Silva, I.; Lynn, T. Accelerometer-Based Human Fall
Detection Using Convolutional Neural Networks. Sensors 2019, 19, 1644. [CrossRef]
33. Wang, G.; Li, Q.; Wang, L.; Zhang, Y.; Liu, Z. Elderly Fall Detection with an Accelerometer Using
Lightweight Neural Networks. Electronics 2019, 8, 1354. [CrossRef]
34. Jahanjoo, A.; Tahan, M.N.; Rashti, M.J. Accurate fall detection using three-axis accelerometer sensor and
MLF algorithm. In Proceedings of the 2017 3rd International Conference on Pattern Recognition and Image
Analysis (IPRIA), Shahrekord, Iran, 19–20 April 2017; pp. 90–95. [CrossRef]
35. Ramachandran, A.; Adarsh, R.; Pahwa, P.; Anupama, K.R. Machine Learning-based Fall Detection
in Geriatric Healthcare Systems. In Proceedings of the 2018 IEEE International Conference on Advanced
Networks and Telecommunications Systems (ANTS), Indore, India, 16–19 December 2018; pp. 1–6.
[CrossRef]
36. Wisesa, I.W.W.; Mahardika, G. Fall detection algorithm based on accelerometer and gyroscope sensor data
using Recurrent Neural Networks. IOP Conf. Ser. Earth Environ. Sci. 2019, 258, 012035. [CrossRef]
37. Hussain, F.; Hussain, F.; Ehatisham-ul Haq, M.; Azam, M.A. Activity-Aware Fall Detection and Recognition
Based on Wearable Sensors. IEEE Sens. J. 2019, 19, 4528–4536. [CrossRef]
38. Musci, M.; De Martini, D.; Blago, N.; Facchinetti, T.; Piastra, M. Online fall detection using recurrent neural
networks. arXiv 2018, arXiv:1804.04976.
39. Liu, K.C.; Hsieh, C.Y.; Hsu, S.J.P.; Chan, C.T. Impact of Sampling Rate on Wearable-Based Fall Detection
Systems Based on Machine Learning Models. IEEE Sens. J. 2018, 18, 9882–9890. [CrossRef]
40. Krizhevsky, A.; Sutskever, I.; Hinton, G.E. Imagenet classification with deep convolutional neural networks.
Advances in Neural Information Processing Systems; AcM: New York, NY, USA, 2012; pp. 1097–1105.
41. Deng, J.; Dong, W.; Socher, R.; Li, L.J.; Li, K.; Fei-Fei, L. ImageNet: A Large-Scale Hierarchical Image
Database. In Proceedings of the 2009 IEEE Conference on Computer Vision and Pattern Recognition, Miami,
FL, USA, 20–25 June 2009.
42. He, K.; Zhang, X.; Ren, S.; Sun, J. Deep residual learning for image recognition. In Proceedings of the IEEE
Conference on Computer Vision and Pattern Recognition, Chengdu, China, 15–17 December 2017; pp. 770–778.
43. Wang, Z.; Yan, W.; Oates, T. Time Series Classification from Scratch with Deep Neural Networks: A Strong
Baseline. CoRR 2016, abs/1611.06455. Available online: https://arxiv.org/abs/1611.06455 (accessed on
30 October 2020).
44. Rumelhart, D.E.; Hinton, G.E.; Williams, R.J. Learning representations by back-propagating errors. Nature
1986, 323, 533–536. [CrossRef]
Technologies 2020, 8, 72 17 of 17

45. Hochreiter, S.; Schmidhuber, J. Long short-term memory. Neural Comput. 1997, 9, 1735–1780. [CrossRef]
46. Virtanen, P.; Gommers, R.; Oliphant, T.E.; Haberland, M.; Reddy, T.; Cournapeau, D.; Burovski, E.;
Peterson, P.; Weckesser, W.; others. SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python.
Nat. Methods 2020, 17, 261–272. [CrossRef]
47. Sztyler, T.; Stuckenschmidt, H. On-body Localization of Wearable Devices: An Investigation of
Position-Aware Activity Recognition. In Proceedings of the 2016 IEEE International Conference on
Pervasive Computing and Communications (PerCom), IEEE Computer Society, Sydney, Australia, 14–19
March 2016; pp. 1–9. [CrossRef]
48. Goodfellow, I.; Bengio, Y.; Courville, A. Deep Learning; MIT Press: Cambridge, MA, USA. 2016.
49. Um, T.T.; Pfister, F.M.J.; Pichler, D.; Endo, S.; Lang, M.; Hirche, S.; Fietzek, U.; Kulić, D. Data
augmentation of wearable sensor data for parkinson’s disease monitoring using convolutional neural
networks. In Proceedings of the 19th ACM International Conference on Multimodal Interaction, Glasgow,
Scotland, 13–17 November, 2017; Lank, E., Ed.; ACM: New York, NY, 2017; pp. 216–220. [CrossRef]
50. Paszke, A.; Gross, S.; Chintala, S.; Chanan, G.; Yang, E.; DeVito, Z.; Lin, Z.; Desmaison, A.; Antiga, L.;
Lerer, A. Automatic differentiation in PyTorch. In Proceedings of the NIPS 2017 Workshop Autodiff
Submission, Long Beach, CA, USA, 9 December, 2017.
51. Kingma, D.P.; Ba, J. Adam: A method for stochastic optimization. arXiv 2014, arXiv:1412.6980.
52. He, K.; Zhang, X.; Ren, S.; Sun, J. Delving deep into rectifiers: Surpassing human-level performance on
imagenet classification. In Proceedings of the IEEE International Conference on Computer Vision, Santiago,
Chile, 7–13 December 2015; pp. 1026–1034.
53. Springenberg, J.T.; Dosovitskiy, A.; Brox, T.; Riedmiller, M. Striving for simplicity: The all convolutional net.
arXiv 2014, arXiv:1412.6806.
54. Fawaz, H.I.; Forestier, G.; Weber, J.; Idoumghar, L.; Muller, P. Deep learning for time series classification:
A review. CoRR 2018, abs/1809.04356. Available online: https://arxiv.org/abs/1809.04356 (accessed on 30
October 2020).
55. Wu, J.; Leng, C.; Wang, Y.; Hu, Q.; Cheng, J. Quantized convolutional neural networks for mobile devices.
In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA,
27–30 June 2016; pp. 4820–4828.
56. Zhou, B.; Khosla, A.; Lapedriza, A.; Oliva, A.; Torralba, A. Learning deep features for discriminative
localization. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas,
NV, USA, 27–30 June 2016; pp. 2921–2929.
57. Aadahl, M.; Andreasen, A.H.; Hammer-Helmich, L.; Buhelt, L.; Jørgensen, T.; Glümer, C. Recent temporal
trends in sleep duration, domain-specific sedentary behaviour and physical activity. A survey among
25–79-year-old Danish adults. Scand. J. Public Health 2013, 41, 706–711. [CrossRef]

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional
affiliations.

c 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).