Page 39 - Read Online

Page 16 of 34 Ma et al. Soft Sci 2024;4:26 https://dx.doi.org/10.20517/ss.2024.20

illustrating a multifunctional wireless platform with different layers of inertial, temperature, humidity, and breathing sensors. Reproduced
with permission [98] . Copyright 2022, American Chemical Society; (E) Schematic diagram of an electric glove integrated with pressure,
temperature, ECG, and humidity sensors. Reproduced with permission [72] . Copyright 2023, American Chemical Society; (F) Schematic
illustrating layered structures of multifunctional sensors. Reproduced with permission [74] . Copyright 2022, Springer Nature; (G)
Illustration of stretchable LIG/hydrogel-based multifunctional wearable electronics; (H) An optical image of developed LIG-based
sensing multimodal; (I) Real-time synchronous monitoring of typical five indicators at normal, walking, and jogging states. Reproduced
with permission [27] . Copyright 2023, Springer Nature. LIG: Laser-induced-graphene; ECG: electrocardiography.

They benefited from conformal features; ultrathin multifunctional skin electronics achieve the perfect
balance between sensing performance and user comfort level. As mentioned above, Lu et al. proposed a
creative strategy, termed the cryogenic transfer approach, completing the transfer of LIG onto ultrathin
(~11 μm in thickness) PPH/PDMS substrate . They designed a stretchable and soft skin-integrated sensing
[27]
patch to collect biophysical signals containing mechanical, temperature, humidity, and ECG [Figure 8G and
H]. A standalone multimodal skin electronics system was established by integrating the fabricated soft
sensor system with a flexible printed circuit board (fPCB) for signal processing and transmission. Five
indicators, including respiration rate, ECG, heart rate, skin temperature and humidity, were collected from
the volunteer performing three activities (baseline, walking, and jogging) [Figure 8I]. As expected, all these
five detected indicators reflected the corresponding health status. It was clear that this multimodal skin
electronics system offered a viable prediagnostic strategy for intelligent healthcare monitoring. Typical LIG-
based multimodal biophysical sensors for healthcare are summarized in Table 2.

2
Machine learning-assisted LIGS E
The fast development of machine learning benefits the design and fabrication of flexible electronics,
facilitates signal processing, and improves system performance [99,100] . Lu et al. developed a TENG-based soft
tactile sensor system, where machine learning was employed to guide the design, including output signals
selection and fabrication parameters . The proposed tactile sensor consisted of four layers, i.e., top and
[101]
bottom PDMS encapsulations, porous LIG-based interdigital electrodes, and a Fluorinated ethylene
propylene (FEP) film [Figure 9A]. Its working mechanism was described as follows: When the human
finger touched the PDMS surface, the skin was positively charged, originating from their different electron
affinities. Following this, horizontal sliding of the charged finger will generate electron transfer between the
LIG electrodes, originating from the electrostatic induction principle [Figure 9B]. With the assistance of
machine learning for six contact modalities analysis, the parameter values of the output signals, the
distribution density of LIG electrodes, and diverse surface microstructures were optimized. The developed
tactile sensor with optimal sensing performance could precisely distinguish ten braille numbers with a high
accuracy of 96.12% by utilizing a customized convolutional neural network (CNN) model [Figure 9C].

Meanwhile, Xie et al. proposed a machine learning-assisted soft pressure sensor based on LIG to realize
real-time tactile perception and voice recognition [Figure 9D] . The intelligent pressure sensor integrated
[102]
with a triboelectric layer converted pressure stimulus into electrical signals, benefiting from the contact
electrification effect. Due to its softness and high performance, the developed intelligent pressure sensor
could be attached to a facemask to collect signal outputs that responded to the volunteer’s speaking. The
intelligent pressure sensor could recognize distinct voices when the speaker pronounced three different
sentences: “How do you do?”, “Nice to meet you”, and “See you later” [Figure 9E]. With the help of
machine learning, the intelligent pressure sensor could accurately classify various voices, achieving a high
accuracy of approximately 94.6% [Figure 9F].

The soft pulse sensor exhibits great potential for real-time monitoring and intelligent management of
cardiovascular health. Most existing soft pulse sensors have limitations, including high cost, clinical