Page 177 - Read Online

Arab Hassani. Soft Sci 2023;3:31 https://dx.doi.org/10.20517/ss.2023.23 Page 25 of 33

demonstrates the SCMN attached as an artificial skin onto a wooden hand with and without sensing area
expansion. This demonstration proved the adjustability and expandability of the system. In addition, the
SCMN exhibited stable performance under 300% expansion, as shown in the pressure mappings obtained
under an applied pressure load. The SCMN could be in applications related to health monitoring,
humanoid robotics and prosthetics, and human-machine interfaces.

CONCLUSION AND FURTHER PERSPECTIVES
The use of single sensors that can detect bio-signals in healthcare applications leads to false-positive or
[163]
negative results . Drawing inspiration from the multitude of bioreceptors that occur naturally in groups of
millions within the body, soft sensor arrays have been developed to provide much more meaningful, robust,
and reliable results. However, the development of flexible, stretchable, and conformable sensor arrays
requires careful selection of durable and biocompatible materials [164,165] . When improving the flexibility and
stretchability of these arrays, one should not compromise their sensitivity and resolution [81,166] . Adhesion,
biodegradability, permeability, and water repellence are the other material factors that should be considered
[167]
based on the target application . The manufacturing techniques used to fabricate these arrays should be
scalable, cost-effective, and reliable to facilitate their widespread adoption in various applications [168,169] . The
development of sensors in the array format provides high spatial resolution, but crosstalk and interference
[170]
between sensors should be avoided to ensure accurate signal detection . Moreover, the signals obtained
using sensor arrays may contain abundant information from untargeted stimuli, and therefore, it is critical
to decouple this interference . To ensure long-term detection stability of sensor arrays under complex
[171]
application conditions, one must carefully select suitable encapsulation materials, forms, and methods, an
[172]
area that warrants further investigation . The interfacing of soft arrays with rigid electronics is another
challenge that calls for the development of flexible and compatible interconnects .
[173]
The complex signals recorded by sensor arrays often necessitate the use of advanced algorithms and
machine learning (ML) techniques to realise real-time signal processing, noise reduction, and data
interpretation [174,175] . The selection of appropriate algorithms to process different types of raw data is critical
to achieve a correct and robust outcome. For instance, to identify various objects using a pressure sensor
array, a multilayer perceptron (MLP) (i.e., a class of ANNs) can be used. MLP is effective for classification
without explicit models because the shapes, sizes, and materials of the target objects are not known in
advance. A convolution neural network (CNN) is another algorithm that can be used in conjunction with
pressure sensor arrays for object detection. For e-nose arrays, PCA (i.e., a typical unsupervised ML
algorithm) is widely used to implement feature clustering and classification.

The signal generated during the human perception process through various receptors is pre-processed in
different transfer stages, and the process is completed in the brain. The unique sensory abilities of individual
receptors, their selective responses, and signal processing at multiple levels ease the workload of the brain.
Furthermore, the transmission of information through diverse channels at varying speeds facilitates the
simultaneous processing of diverse types of data . Similarly, in the bioinspired systems, we need to use
[176]
materials with distinct mechanical characteristics, sensitivity, and selectivity to different stimuli. This fact
has inspired several researchers to develop multimodal soft sensor arrays. However, mechanical isolation of
various types of sensors should be attempted to minimise crosstalk and interference, as well as to improve
their selectivity. Moreover, considering the different response times of different sensors could help trigger
them one by one for processing to limit inter-sensor interference. Finding a trade-off between spatial
resolution and interference for sensor arrays is an extremely difficult endeavour. The spatial resolution
varies at the locations of different organs in the human body. Moreover, the spatial resolution can vary
depending on the type of receptor. If the developed bioinspired systems can follow these categorisations, we