Conference PaperPDF Available

Abstract and Figures

Sensor calibration is vital to have valid measurements of physical activities. Herein we deal with adjusting the signal from a wearable force sensor against a reference scale. By using a few samples and data augmentation, we trained a neural-based regression model to correct the wearable output. For this task, we tested the novel Auto-Rotating Perceptrons (ARP). We found that a neural ARP model with sigmoid activations can outperform an identical neural network based on classic per-ceptrons with sigmoid and even ReLU activation.
Content may be subject to copyright.
Smart Sensor Calibration with Auto-Rotating Perceptrons
Daniel Saromo 1 2 Leonardo Bravo 2Elizabeth Villota 2
Sensor calibration is vital to have valid measure-
ments of physical activities. Herein we deal with
adjusting the signal from a wearable force sensor
against a reference scale. By using a few samples
and data augmentation, we trained a neural-based
regression model to correct the wearable output.
For this task, we tested the novel Auto-Rotating
Perceptrons (ARP). We found that a neural ARP
model with sigmoid activations can outperform
an identical neural network based on classic per-
ceptrons with sigmoid and even ReLU activation.
1. Introduction
In recent years, wearable technology has seen an increase
in sports applications. Unlike conventional motion capture
systems, wearables can gather data outside the laboratory
environments (Adesida et al.,2019). Sensing of ground re-
action forces (GRF) in particular is of great interest for per-
formance evaluation in sports (Evans et al.,2018). Wearable
technology for GRF sensing mainly consists of insole-type
wearables that measure plantar pressure distribution (Naga-
mune & Yamada,2018). Calibration of these devices has
been performed with classical statistical approaches (Eguchi
& Takahashi,2018) and also using neural networks (Eng
et al.,2018). For instance, an LMBP-ANN model served to
predict validated load signals from a nonlinear pressure sen-
sor (Almassri et al.,2018), a WNN model predicted GRFs
using plantar information from insole pressure sensors (Sim
et al.,2015), and an LSTM model predicted the strain state
from a stretch sensor (Oldfrey et al.,2019).
This paper presents the calibration of an insole-type GRF
sensor using the recently developed Auto-Rotating Percep-
trons (ARP) for artificial neural networks. We aim to im-
plement and train a feedforward network model in order
PUCP Artificial Intelligence Research Group
PUCP Biome-
chanics and Applied Robotics Research Laboratory. Correspon-
dence to: Daniel Saromo
, Leonardo
, Elizabeth Villota
LXAI Workshop at
37 th
International Conference on Machine
Learning, Vienna, Austria, PMLR 108, 2020. Copyright 2020 by
the author(s).
to have a valid posterior measurement without requiring
reference signals.
2. Problem definition
A WEarable VErtical GRF Sensor system (WEVES) com-
posed of an insole with piezoelectric sensors and an elec-
tronic housing -to be attached to the leg- was developed,
see Figure 1(a). WEVES was designed to measure forces
up to
2600 N
. A 65-
male amateur athlete wore WEVES
and performed a standardized movement on an AMTI force
plate (OPT400600 model), see Figure 1(b). AMTI and
WEVES signals were recorded for
30 s
, which we show in
Figure 1(c). Even though AMTI and WEVES signals are
quite similar in shape, there exists a difference in scale, and
this is recurrent in all sample pairs recorded. Let
be the
scaling factor between AMTI and WEVES measurements.
Then, WEVES calibration consists in finding that factor,
which varies across the signal pairs.
(a) (b) (c)
Figure 1.
(a) WEVES mechatronic system: electronic box and
insole. (b) Stand-up straight position test with AMTI (red) and
WEVES (green) sensors. (c) Force measurements (
(red), uncalibrated WEVES (green), and scaled WEVES (blue).
3. Methodology
WEVES calibration was performed employing a neural re-
gression model that predicts the
factor. We compared
four model types with the same architecture, half of them
had classic perceptrons, and the others used the novel ARP
units (Saromo et al.,2019). ARP and classical perceptrons
were tested for both sigmoid and ReLU activation func-
tions. Previous ARP implementations have reported that
this neural unit can lead to better performance in image
classification tasks, provided that the sigmoid is the network
activation function (Saromo et al.,2019).
Smart Sensor Calibration with Auto-Rotating Perceptrons
Dataset preparation:
In order to find the target
, we need
to quantify the similarity between AMTI platform signal and
filtered WEVES outputs. These signals are represented by
the vectors
, respectively. We define the root mean
square error (RMSE) as the loss function, so that finding the
Kfactor can be posed as an optimality problem:
K= arg min
{RMSE (p, Kw)},K > 0.
Several algorithms find the best solution for such an op-
timization problem (Karaboga,2005;Wang et al.,2018;
Kumari & Kaur,2019). We have implemented the Artificial
Bee Colony (ABC) and the Particle Swarm Optimization
(PSO) algorithms to obtain
values for each signal pair
). The
factor chosen corresponds to the lowest cost
function value found. Descriptive statistics and information
related to these experiments are shown in Table A1.
WEVES signals were obtained at
100 Hz
and then interpo-
lated to reach
1 MHz
, which is the frequency of the AMTI
output data. We generated the
values and then downsam-
pled both signals by a factor of 15. Hence, we reduced the
input data dimension to 2000. We then filtered WEVES
signals by using a 4th-order low-pass Butterworth filter with
a normalized cutoff frequency of 0.15, as implemented in
literature for these signal types (Pandit et al.,2018). Data
was then circularly shifted, retaining important information
-the signal’s middle part, see Figure 1(c). By this artifice, the
number of pair samples increased from 12 to 8787. Result-
ing signals were further scaled to the range
. The final
dataset was divided into three groups: 7029 for training, 879
for validation, and the remaining 829 for testing.
Architecture definition:
Aiming for a real-time implemen-
tation, we chose a feedforward neural network with only one
hidden layer and the following architecture: 2000-256-1.
The scalar output is the scaling factor
that the regression
neural network needs to predict. We assigned the same
initial weights and biases to the four network types before
each execution. For the experiments, the ARP hyperparam-
eters are defined as:
xQ= 2.6
, since the maximum input
value is
1< xQ
; and
L= 3
, because the unipolar sigmoid
derivative σ0(z)is not very small for |z| ≤ 3.
4. Experiments: Classic perceptrons vs ARP
Figure 2shows the loss evolution throughout the epochs.
We observe that the networks with vanilla units activated
by sigmoids rapidly stagnate in a constant loss value. This
undesired premature convergence happens for both the train
and validation losses. By far, this is the worst-performing
model type. The other model family, which uses classic
perceptrons with ReLU, presents a loss decay consistent
over time and performs better than the first model type. The
following two model groups had ARP in their hidden layer
neurons. The ReLU-activated ARP model family exhibit
an initial better performance than the other ReLU network
type with classical perceptrons, but then is overtaken. The
remaining two curves show information of the last model set,
an ARP network with sigmoid activations. This model type
has the best performance, reaching the lowest loss values,
even when compared to the ReLU networks. Notably, with
ARP, the test loss of the sigmoid networks was reduced by
a factor of 15 at the cost of increasing the execution time by
12% (for details see Figure A1).
Figure 2.
Loss function of the four model types. Central bold
lines are median curves, and shades represent interquartile ranges.
Number of executions per model type: 50. Epochs per iteration:
100. Batch size: 64. Optimizer: Nadam (lr = 0.003).
5. Conclusions
In this work, we employed the ARP neural unit aiming to
calibrate a wearable GRF sensor. We compared the ARP
performance against models with classic perceptrons using
two different activation functions. The results show that
in a neural network with sigmoid activations, changing the
classic perceptrons to ARP can lead to a better performance,
even beating the same neural model with vanilla perceptrons
and ReLU activations. Also, even though the ARP units
were designed to be employed with saturated functions, we
can see that ARP can improve the model performance for
the early epochs, when using them with a non-saturated
function like ReLU. A more in-depth analysis needs to be
performed to explain this behavior appropriately.
The calibration pipeline herein presented could be extrapo-
lated to other cases where it is required to calibrate a sensor
output to make it close to a reference scaled signal.
Smart Sensor Calibration with Auto-Rotating Perceptrons
This work is funded by CONCYTEC-FONDECYT Peru
(contract number E041-01/058-2018-FONDECYT-BM-
Adesida, Y., Papi, E., and McGregor, A. Exploring the
role of wearable technology in sport kinematics and
kinetics: A systematic review. Sensors, 2019. doi:
Almassri, A., Wan Hasan, W., Ahmad, S., Shafie, S., Wada,
C., and Horio, K. Self-calibration algorithm for a pressure
sensor with a real-time approach based on an artificial
neural network. Sensors, 18(8):2561, 2018. doi: 10.3390/
Eguchi, R. and Takahashi, M. Accessible calibration of in-
sole force sensors using the wii balance board for kinetic
gait analysis. In 2018 IEEE SENSORS, pp. 1–4, 2018.
Eng, S., Al-Mai, O., and Ahmadi, M. A 6 dof, wearable,
compliant shoe sensor for total ground reaction measure-
ment. IEEE Transactions on Instrumentation and Mea-
surement, 67(11):2714–2722, 2018.
Evans, M., Colyer, S., Cosker, D., and Salo, A. Foot contact
timings and step length for sprint training. In Proceedings
of the 2018 IEEE Winter Conference on Applications of
Computer Vision (WACV 2018), 03 2018. doi: 10.1109/
Karaboga, D. An idea based on honey bee swarm for numer-
ical optimization. Technical report, Erciyes University,
Kumari, K. and Kaur, P. Use of nature-inspired computing
techniques in real world applications (2010-2019) a brief
review. International Journal of Computer Applications
& Information Technology, 11(2):258–264, 2019.
Nagamune, K. and Yamada, M. A wearable measurement
system for sole pressure to calculate center of pressure
in sports activity. In Proceedings of the 2018 IEEE Inter-
national Conference on Systems, Man, and Cybernetics
(SMC 2018), pp. 1333–1336, 10 2018.
Oldfrey, B., Jackson, R., Smitham, P., and Miodownik, M.
A deep learning approach to non-linearity in wearable
stretch sensors. Front. Robot. AI 6: 27. doi: 10.3389/frobt,
Pandit, S., Godiyal, A., Vimal, A., Singh, U., Joshi, D., and
Kalyanasundaram, D. An affordable insole-sensor-based
trans-femoral prosthesis for normal gait, 2018.
Saromo, D., Villota, E., and Villanueva, E. Auto-Rotating
Perceptrons. LatinX in AI Workshop at NeurIPS 2019
(arXiv:1910.02483), 2019.
Sim, T., Kwon, H., Oh, S., Joo, S.-B., Choi, A., Heo, H.,
Kim, K., and Mun, J. Predicting complete ground reaction
forces and moments during gait with insole plantar pres-
sure information using a wavelet neural network. Journal
of biomechanical engineering, 137, 06 2015.
Wang, D., Tan, D., and Liu, L. Particle swarm optimization
algorithm: an overview. Soft Computing, 22(2):387–408,
A. Supplementary Material
Table A1.
Comparison of the algorithms used to calculate
. Both
algorithms ran 100 times with 50 elements (food sources and
particles for the ABC and PSO algorithms, respectively).
MEA N 32.712 29.440
STD . DE V. 7.458 7.159
TI ME [s]
MEA N 31.908 3.121
STD . DE V. 0.206 0.059
Figure A1.
Best test loss value (left vertical axis) and correspond-
ing processing time (right vertical axis) per execution. Each one
of the executions has 100 epochs. The loss function is the loga-
rithm of the hyperbolic cosine of the network prediction error. The
bars represent the median value obtained after 50 executions. The
vertical lines at the top of the bars represent the interquartile range.
ResearchGate has not been able to resolve any citations for this publication.
Conference Paper
Full-text available
This paper proposes an improved design of the perceptron unit to mitigate the vanishing gradient problem. This nuisance appears when training deep multilayer perceptron networks with bounded activation functions. The new neuron design, named auto-rotating perceptron (ARP), has a mechanism to ensure that the node always operates in the dynamic region of the activation function, by avoiding saturation of the perceptron. The proposed method does not change the inference structure learned at each neuron. We test the effect of using ARP units in some network architectures which use the sigmoid activation function. The results support our hypothesis that neural networks with ARP units can achieve better learning performance than equivalent models with classic perceptrons.
Full-text available
Nature-inspired computing techniques are the emerging computing techniques that have been used in a wide sphere of applications. The study presents a brief review of some of the major applications of these techniques. This study is restricted to the manuscripts which were published during 2010-2019 only. It is not possible to cover a large number of applications in one single brief study. Therefore, here, four different types of applications such as disease diagnosis, query optimization, sentiment analysis and stock data mining have been only considered.
Full-text available
The aim of this review was to understand the use of wearable technology in sport in order to enhance performance and prevent injury. Understanding sports biomechanics is important for injury prevention and performance enhancement and is traditionally assessed using optical motion capture. However, such approaches are limited by capture volume restricting assessment to a laboratory environment, a factor that can be overcome by wearable technology. A systematic search was carried out across seven databases where wearable technology was employed to assess kinetic and kinematic variables in sport. Articles were excluded if they focused on sensor design and did not measure kinetic or kinematic variables or apply the technology on targeted participants. A total of 33 articles were included for full-text analysis where participants took part in a sport and performed dynamic movements relating to performance monitored by wearable technologies. Inertial measurement units, flex sensors and magnetic field and angular rate sensors were among the devices used in over 15 sports to quantify motion. Wearable technology usage is still in an exploratory phase, but there is potential for this technology to positively influence coaching practice and athletes’ technique.
Full-text available
This paper presents a novel approach to predicting self-calibration in a pressure sensor using a proposed Levenberg Marquardt Back Propagation Artificial Neural Network (LMBP-ANN) model. The self-calibration algorithm should be able to fix major problems in the pressure sensor such as hysteresis, variation in gain and lack of linearity with high accuracy. The traditional calibration process for this kind of sensor is a time-consuming task because it is usually done through manual and repetitive identification. Furthermore, a traditional computational method is inadequate for solving the problem since it is extremely difficult to resolve the mathematical formula among multiple confounding pressure variables. Accordingly, this paper describes a new self-calibration methodology for nonlinear pressure sensors based on an LMBP-ANN model. The proposed method was achieved using a collected dataset from pressure sensors in real time. The load cell will be used as a reference for measuring the applied force. The proposed method was validated by comparing the output pressure of the trained network with the experimental target pressure (reference). This paper also shows that the proposed model exhibited a remarkable performance than traditional methods with a max mean square error of 0.17325 and an R-value over 0.99 for the total response of training, testing and validation. To verify the proposed model’s capability to build a self-calibration algorithm, the model was tested using an untrained input data set. As a result, the proposed LMBP-ANN model for self-calibration purposes is able to successfully predict the desired pressure over time, even the uncertain behaviour of the pressure sensors due to its material creep. This means that the proposed model overcomes the problems of hysteresis, variation in gain and lack of linearity over time. In return, this can be used to enhance the durability of the grasping mechanism, leading to a more robust and secure grasp for paralyzed hands. Furthermore, the exposed analysis approach in this paper can be a useful methodology for the user to evaluate the performance of any measurement system in a real-time environment.
Full-text available
This paper proposes a novel and an affordable lower limb prosthesis to enable normal gait kinematics for trans-femoral amputees. The paper details the design of a passive prosthesis with magneto-rheological (MR) damping system and electronic control. A new control approach based on plantar insole feedback was employed here. Strategically placed sensors on the plantar insole provide required information about gait cycle to a finite state controller for suitable action. A proportional integral (PI) based current controller controls the required current for necessary damping during gait. The prosthesis was designed and developed locally in India keeping in view the cost, functionality, socio-economic, and aesthetic requirements. The prototype was experimentally tested on a trans-femoral amputee and the results are presented in this work. The implementation of the proposed design and control scheme in the prototype successfully realizes the notion that normal gait kinematics can be achieved at a low cost comparable to passive prostheses. The incurring cost and power expenditure of the proposed prosthesis are evaluated against passive and active prostheses, respectively. The commercial implications for the prosthesis were explored on the basis of recommendations of ISPO Consensus Conference on Appropriate Prosthetic Technology in Developing Countries. The key objective of this work is to enable lucid design for development of an affordable prosthesis in a low-resource setting.
Full-text available
Particle swarm optimization (PSO) is a population-based stochastic optimization algorithm motivated by intelligent collective behavior of some animals such as flocks of birds or schools of fish. Since presented in 1995, it has experienced a multitude of enhancements. As researchers have learned about the technique, they derived new versions aiming to different demands, developed new applications in a host of areas, published theoretical studies of the effects of the various parameters and proposed many variants of the algorithm. This paper introduces its origin and background and carries out the theory analysis of the PSO. Then, we analyze its present situation of research and application in algorithm structure, parameter selection, topology structure, discrete PSO algorithm and parallel PSO algorithm, multi-objective optimization PSO and its engineering applications. Finally, the existing problems are analyzed and future research directions are presented.
Full-text available
In general, three-dimensional ground reaction forces (GRFs) and ground reaction moments (GRMs) occurring during human gait motion are measured using a force plate, which has an expensive price and a spatial limitation. Therefore, we proposed in this study, a prediction model for GRFs and GRMs using only plantar pressure information measured from an insole pressure sensor. For this, the prediction model estimated GRFs and GRMs for a group of adolescent idiopathic scoliosis (AIS) patients (N = 10), a group of normal healthy subjects (N = 10) and for the total group (N = 20) comprised of both AIS patients and healthy subjects. Model performance was validated using correlation coefficients (r) and the normalized root mean square (NRMSE) and was compared to the prediction accuracy of the previous methods using the same data set. As a result, the performance of the GRF and GRM prediction model proposed in this study (Normal group: r = 0.847-0.988, NRMSE = 10.920-19.216%; Patient group: r = 0.822-0.990, NRMSE = 9.170-15.114%; Total group: r = 0.850-0.989, NRMSE = 10.117-17.182%) was better than that of the previous linear method and non-linear methods for every group and component (p < 0.05 or 0.01). The results indicate that the proposed three-dimensional GRF and GRM prediction model has improved performance, overcomes the spatial limitation of the force plate, and can be applied in various fields to allow single-model analysis without conducting optimal model selection processes.
In this paper, we present a compliant six-axis sensor for total ground reaction measurement, the calibration methodology, and preliminary data. The sensor is intended for gait analysis and is designed to have minimal effect on natural gait. A triaxial optical force sensor is combined with an array of pressure sensing films to form a wearable compliant six-axis force/moment sensor. Two sensor units were developed for the toe and heel and used in two types of experiments: stepping on the sensors that are mounted on the ground and attaching the sensors under the shoe while walking. The data from the sensors are compared with measurements obtained from a standard force plate. The deflections induced by the sensor compliance exhibit a slight nonlinear force-deflection relation. Regardless of the nonlinear effects, the sensor is accurately calibrated with a linear least squares method. To see how well these nonlinearities could be calibrated for, a nonlinear calibration with a neural network was used. For the sensors attached to the floor, a linear calibration achieved an RMSE of 4.49% while the neural network achieves 2.68%. For the wearable sensor, the linear calibration RMSE was 9.39% and the neural network RMSE was 5.21%. When the orientation data of the sensors (measured by a motion capture system) were added as an input to the neural network calibration, the RMSE was reduced to 3.25%.
Conference Paper
The frequency and length of a runner's steps are fundamental aspects of their performance. Accurate measurement of these parameters can provide valuable feedback to coaching staff, particularly if regular measurement can be made and monitored over the course of a season. This paper presents a computer vision based approach using high framerate cameras to measure the location and timing of foot contacts from which step length and frequency can be determined. The approach is evaluated against force-plates and optical motion capture for a mix of 18 trained and recreational runners. Force-plates and optical motion capture are considered to be the current "gold-standard'' in biomechanics, and this is the first vision based paper to evaluate against these standards. Landing and take-off times were shown to be measurable to within 1.5 frames (at 180fps) and step length to within 1 cm.