A Pathway into Metaverse: Gesture Recognition Enabled by Wearable Resistive Sensors

Hand gesture recognition is of great importance for human–Metaverse interaction. The human hand is relatively smaller, with very complex articulations than the entire human body. Therefore, precise hand gesture recognition, especially in complicated environments, is still a great challenge. Wearable resistive sensors that could directly characterize joint movements are one of the most promising technologies for hand gesture recognition due to their easy integration, low cost, and simple signal acquisition. Here, we summarize common categories of wearable resistive sensors for hand gesture recognition and review the recent advances. Besides, common strategies for improving waterproofness (molecular design, material modification, and interfacial encapsulation) are also outlined. Perspectives on the advantages of combining machine vision and resistive sensors for accurate hand gesture recognition and dynamic hand gesture tracking are then provided. We hope this review could inspire future research in wearable technologies for human–Metaverse systems.


Introduction
Metaverse was coined for the first time by Neal Stephenson in his book Snow Crash. At present, Metaverse is defined as an open, shared, and persistent 3D virtual world where digital avatars and reality co-exist, interact together, and go beyond physical, mental, or economic barriers. [1][2][3][4] The functioning of Metaverse requires a combination of cutting-edge technologies like blockchain, augmented reality, virtual reality, 3D reconstruction, artificial intelligence, the Internet of Things, and interactive technologies. [3] From the perspective of the sense of reality, interactive technologies are the fundamental enablers for accessing an immersive DOI: 10.1002/adsr.202200054 and realistic experience for users in the Metaverse. As a bridge between the physical and virtual worlds, the interactive technologies could not only collect information from the real world and project the information into the Metaverse but also feed information from the virtual world back to the real world through haptic devices. [5][6][7][8][9][10] Hand gesture recognition technologies could collect the posture and action of human hands and control their avatars to interact with other users or nonplayer characters (NPCs) for providing interactive and immersive experiences assisted by virtual reality. [11] In industry, hand gestures are often recognized by visualbased techniques, using machine vision algorithms to process hand-related visual images. [12] However, visual-based approaches are costly and not energy efficient. The accuracy is also limited by the qualities of images, which could be greatly affected by environmental interference and varied light conditions. [13,14] To minimize the influence of environmental conditions on hand gesture recognition, wearables or skinattachable sensors that could directly detect joint movement gradually become a hot spot. [15][16][17][18][19] Among these sensors, resistive sensors have attracted much attention in the scientific community owing to their simple structure, easy integration, low cost, and simple signal acquisition. [20][21][22][23][24][25] Yet, since the joint motion could cyclically produce strain up to 70% within a short time (hundreds of microseconds), [26,27] the performance index of resistive sensors should be met to accurately capture human posture, including stretchability, cyclic stability, gauge factor, and response time. Besides, wearable sensors are hard to avoid being exposed to moisture or even watery environment in daily activities. Waterproofness of wearable resistive sensors is also highly required. [28] In this review, we focus on the recent advances in hand gesture recognition based on wearable resistive sensors. The sensing mechanisms of resistive sensors could be divided into three categories including 1) monitoring the normal pressure at the convex surface of joints induced by jointing bending; 2) detecting the tensile strain at the convex surface of joints; 3) utilizing the compression strain at the concave surface of joints during bending. We also summarize and compare the state-of-art resistive sensors for posture recognition according to the performance index of stretchability, cyclic stability, sensitivity, and response time. Next, we emphasize the importance of waterproofness and www.advancedsciencenews.com www.advsensorres.com environmental robustness for resistive sensor-based wearable systems to enhance practical application stability in natural environments. We then discuss the advantages of combining machine vision and resistive sensors for hand gesture recognition and dynamic gesture tracking (Figure 1). We hope this review could inspire future research in wearables for gesture recognition and its application in human Metaverse interactions.

Resistive Sensors for Hand Gesture Recognition
Nowadays, new materials and sensing techniques have been developed and could enable more intuitive and comfortable wearable or skin-attachable sensors for hand gesture recognition. [29][30][31][32][33] Resistive sensors should exhibit advantages in stretchability, sensitivity, response time, and stability so as to detect joint movements. For example, the tensile strain, defined as ΔL/L 0 × 100%, needs to surpass 70%. For resistive strain sensors, the sensitivity is often characterized by gauge factor (GF), defined as ΔR/R 0 ( means strain), which means that the relative resistance changes of sensors under the applied strain. [34] Besides, for resistive pressure sensors, the sensitivity is often defined as (△G/G0)/ (p) (G means the conductance of sensors and is equal to 1/R) and means that the relative conductance changes of sensors under the applied pressure. [35] The response time is the time that the resistance/conductance takes to increase from 10% to 90% of the final electrical signals under an applied stimulus. [30] The stability is often investigated by recording the electrical degradation (such as resistance) of sensors under hundreds or thousands of testing cycles. Through material selection and structural engineering, the performance of resistive sensors could be improved in terms of stretchability, sensitivity, response time, and stability.
In the following sections, We mainly discuss the mechanism of resistive sensors (Figure 2a) and the strategies for performance improvement. The advantages and disadvantages of the three strategies for hand gesture recognition are discussed in Table 1. Then, we summarize the state-of-art resistive sensors in terms of stretchability, sensitivity, response time, and stability, as well as their structural engineering and material selection.

Piezoresistive Sensors
The first strategy is to detect normal pressure at the convex surface induced by joint bending. In this strategy, the piezoresistive sensors employ the mechanism of contact resistance between a rough/structured conductive layer and an electrode ( Figure 2b). [36] Joint bending could induce a normal pressure at the convex surface of joints. This normal pressure can decrease the contact resistance by increasing the contact area between the conductive layer and the conductor. Besides, because the contact resistance is a surface effect, the sensors based on contact  resistance modulation could be very thin, thus enabling good skin comfortability and conformality. [37] Since the piezoresistive pressure sensors need to monitor the normal force at the convex surface of human joints, these sensors should be stretchable, often based on elastomer and textile. [38,39] For example, Liu et al. reported a large-area textile-based pressure sensor by the maskassisted electroless deposition (ELD) method (Figure 3a). [40] The hierarchical, porous nanostructure of both CNT fabric and nanoparticle-coated Ni textiles provide sufficient roughness to induce changes of contact resistance under the applied pressure, contributing to the high sensitivity (14.4 kPa −1 ), low detection limit (2 Pa), and fast response (24 ms). This textile pressure sensor was attached to the index finger and monitored the finger bending. Once the finger is bent, the contact area between the carbon nanotube-coated fabric conductive layer and Ni electrodes increases, thus leading to a resistance decrease. Therefore, the recorded current under the const voltage increased correspondingly (I = U/R). Besides, stretchable elastomers (i.e., polydimethylsiloxane (PDMS), Ecoflex, silicone rubber, styrene ethylene butylene styrene (SEBS)) are often chosen as the matrix for wearables. [41][42][43][44][45] Cheng et al. fabricated an MXene-based piezoresistive sensor with skin-like microspinous microstructures through a simple abrasive paper stencil printing process ( Figure 3b). [46] Then, this sensor was attached to the joints of the human body to monitor wrist flexion and elbow motion ( Figure 3c). The change of contact area under the applied pressure is significantly vital to the sensitivity of pressure sensors. Therefore, structure engineering design in the conductive layer could improve the sensitivity. Duan et al. reported a pressure sensor based on microstructures whose height exhibits the random Gaussian distribution (RGD), which features high sensitivities (2093 kPa −1 ), low limit of detection (< 0.43 mN), and fast response (< 4 ms). [35] Besides, Kong et al. proposed hair-epidermisdermis hierarchical structures to enable the piezoresistive sensor linear sensing range up to 30 kPa and high sensitivity (137.7 kPa −1 ). [47] Other microstructures including pyramid structures, [48] leaf impression, [49] micro-cone arrays, [50] interlocking microstructures, [51] nanowires, [52] micro-rod arrays, [53] micro-sphere arrays, [54] and other bioinspired structures [55] are also fabricated to improve the sensitivity of pressure sensors (Figure 3d-f).  [40] Copyright 2017, Wiley-VCH. b) Fabrication of piezoresistive sensors with bionic spinous microstructure. c) Current responses of wrist bending and elbow swing. b,c) Reproduced with permission. [46] Copyright 2020, American Chemical Society. d-f). Structures on the sensing layer of pressure sensors. d) Reproduced with permission. [54] Copyright 2020, The Royal Society of Chemistry. e) Reproduced with permission. [35] Copyright 2022, Elsevier. f) Reproduced with permission. [55] Copyright 2021, Elsevier.

Crack-Based Resistive Strain Sensors
The second strategy is to detect tensile strain at the convex surface of joints during bending through resistive strain sensors. The resistive strain sensor consists of a strain-sensing layer and two electrodes, supported by a stretchable substrate ( Figure 2c). Two kinds of strain sensor layers are developed to detect the tensile strain. The first is crack-based strain layers. Crack-based strain sensors are sensitive to small strains with high sensitivity. The high strain sensitivity originates from the rare yet large gap-bridging steps on opposite edges of a zigzag crack (Figure 2d). This crack structure is often inspired by the crack-shaped slit organs in spiders or scorpions. [56] For example, inspired by the geometry of the slit organ of spiders, Figure 4. Resistive sensors that detect surface tensile strain. a) Crack-based strain sensor with highly sensitive based on nanoscale crack junctions with a GF of over 2000 between 0 -2% strain. Reproduced with permission. [56] Copyright 2014, Springer Nature. b) Conductive-network-based resistive strain sensors. Reproduced with permission. [57] Copyright 2023, American Chemical Society. c) Liquid-metal-based resistive strain sensors. Reproduced under terms of the CC-BY license. [58] Copyright 2021, Wiley-VCH. d) Hydrogel-based resistive strain sensors. Reproduced under terms of the CC-BY license. [59] Copyright 2021, American Association for the Advancement of Science. e) Ionic-liquid-based resistive strain sensors. Reproduced with permission. [60] Copyright 2020, Wiley-VCH.
Kang et al. developed a strain sensor based on nanoscale crack junctions, which are highly sensitive to strain, with a gauge factor of over 2000 between 0-2% strain (Figure 4a). [56] The electrical conductance of these nano metal strips in the crack-based sensors experiences a sudden jump from a finite value to zero only under a small strain due to the discon-nect of adjacent nano metal strips, thus leading to function failure of crack-based sensors. Therefore, these crack-based sensors could only respond to small strains (< 5%). Most of these crack-based sensors are not applied for human posture monitoring due to the large strain (> 50%) induced by joint movements.

Conductive-Network-Based Resistive Strain Sensors
Strain sensors that can measure large strain are important for the detection of finger-bending behaviors. Compared to the crackbased sensors, the conductive network-based strain sensors could sustain the large strain. In the configuration of conductivenetwork-based resistive strain sensors, the strain-sensitive layer is categorized into conductive-nanomaterial-modified elastomers (Ag, Au, carbon nanotubes, graphene, and MXene), hydrogels, ionic-liquid-based composites, and intrinsically stretchable liquid metal. The stretchability and GF of conductive-network-based resistive strain sensors are highly dependent on composites' mechanical properties, fillers' electrical properties, and structural design.
The elastomers or fabrics as matrix materials often could sustain the large strain, and the distribution of the conductive filling materials could form the percolation conductive network to maintain electrical conduction paths in the elastic composite at large strain. [31] Duan et al. proposed a water-modulated biomimetic hyper-attribute-gel (Hygel) e-skin composed of MXene and silk fibroin (Figure 4b). [57] This hyper-attribute Hygel exhibited high resolution in strain sensing for bending angles, which is eligible for hand motion monitoring and dynamic gesture recognition with long-term wearing comfort, biocompatibility, and safety. This Hygel e-skin was used to achieve 1D-CNN (one-dimensional convolutional neural network) assisted highprecise classification of 15 dynamic gestures.
Because gallium-based liquid metal (gLM) alloys feature low melting points and exhibit fluidic nature, they could retain high electrical conductivity even under extreme stretch conditions and conformally contact with complex surfaces (such as human skin). [61,62] Besides, the fluidic nature endows gLM sensors with low hysteresis. These remarkable characteristics are absent in other conductive materials (silver, gold, carbon, graphene, Mxene, etc.) because of their rigid nature. [58,63] Therefore, gLM alloys are considered highly suitable materials for stretchable resistive strain sensors for human health monitoring and posture detection. As shown in Figure 4c, Wu et al. reported a liquidmetal-based resistive strain sensor via selective wetting and transferring processes, which exhibited an ultralow detection limit (0.05%) and negligible hysteresis. Moreover, This strain sensor incorporated a smart glove and a smart kneecap to detect human gestures. [58] Besides, a liquid-metal-based fiber consisting of liquid metal as the core and the poly(styreneb-(ethylene-cobutylene)-b-styrene) (SEBS) copolymer as the shell was fabricated via the thermal drawing, exhibits a GF of 1.8 and could undergo over 300% strain. [64] Yet, although various printing methods have been developed (such as direct writing, screen printing, inkjet printing, acoustophoretic printing, and 3D printing) to obtain high-resolution liquid metal conductive patterns and sensors on a large scale, some difficulties in high-resolution, high-speed, and large-scale printing techniques need to be resolved. [65] Besides, the stable and robust conductive connection between liquid metal and rigid electronic components is vital for wearable sensor systems. Interfacing the liquid metal with metal electrodes is still challenging due to the alloying behavior of liquid metals with most metals. [28] On the other hand, human skin uses ions and molecules to sense external stimuli. When subjected to external stimuli, the ion channels on the cell membrane open, and Na + in the extracellular matrix flows into the cell and generates electrical signals. Ion-based composite materials that rely on electron-mediated transport and compensation of ions have been recently adopted to mimic the sensing mechanism of human skin to external stimuli. [66,67] The two main ionic materials are hydrogels and ionogels.
Hydrogels exhibit excellent characteristics (flexibility, stretchability, biocompatibility, etc.), fulfilling the demands of wearable sensors for gesture recognition. [68] The conductive network structure of hydrogels deforms upon external strain, so the resistance of hydrogels changes. For example, Qu et al. prepared a ligninreinforced thermoresponsive poly (ionic liquid) hydrogel, which exhibits high stretchability (over 1425%) due to the electrostatic interaction between lignin and ionic liquid (Figure 4d). [59] In spite of high stretchability, this hydrogel sensor presents a relatively low GF (1.37). Various strain-sensitive hydrogels with desired anti-freezing, adhesive, and self-healing capabilities have been widely developed for human activity monitoring. [69][70][71] Yet, more research should be devoted to effectively extending the service life of the hydrogel-based strain sensor, such as improving the mechanical properties of the hydrogel and overcoming the disadvantage of losing water. [57,72] Ionic-liquid-based resistive strain sensors where the sensing layer is often composed of ionic liquid and polymer. For example, Li et al. proposed an ultra-durable ionic resistive strain sensor with excellent healing ability by impregnating ionic liquids into the robust poly(urea-urethane) (PU) network (Figure 4f), which was attached to a finger to detect the finger-bending. [60] Yet, some urgent challenges still need to be solved. For instance, nowadays, research on factors (water content, humidity, temperature, etc.) is not sufficient. More potential models of ionic materials need to be developed so that ionic-liquid resistive sensors could adapt to various complicated environments. [66]

Slit-Based Resistive Compression Sensors
During joint bending, the surface tensile strain on the convex surface often concentrates in the joint and varies unevenly. Besides, the joint folds/creases on the convex surface of joints erect obstacles to mechanical conformality and compliance of wearable sensors. [73] The concave surfaces of joints come close to each other and induce compression strain during joint movements. Therefore, Duan et al. proposed the third strategy to detect joint movement by utilizing the compression strain at the concave surface of joints during bending. [74] Inspired by the geometry and principle of the scorpions' slit sensilla, a bionic wide-range, high-sensitive slit-based MXene motion sensor was developed, consisting of parallel, periodic two-phase components involving MXene-coated polyurethane (PU) hard blocks and soft air gaps (Figure 5a). Such slit structures enable the precise sensation of compression strain on the concave surfaces through a double contact effect during joint movement (Figure 2e). Meanwhile, an equivalent resistance model and corresponding electro circuit of this motion sensor were established to analyze the mechanism (Figure 5b). The contact area among internal MXene-coated PU keel structures and between adjacent MXene-coated PU blocks increases during joint bending and the corresponding bulk resistance (R c ) and interfacial resistance (R b ) decrease, leading to the current increment of the sensor. Compared to other slit ratios, the motion sensor with a 1/2 slit ratio exhibits a high sensitivity of 0.45% deg −1 , and a wide-angle sensing range (15°-120°) (Figure 5c), which could be used to distinguish similar gestures due to the high angle sensitivity (Figure 5d).

Waterproof Sensors
Environmental robustness and waterproofness are also required for resistive sensors since they are hard to avoid being exposed to moist or even watery environments in interactive activities. Most waterproof demonstrations reported so far are based on molecular design, material modification, and encapsulation strategies.
For molecular design, Cao et al. reported an aquatic, stretchable, and self-healing ion gel electronic skin, which could self-heal through highly reversible ion-dipole interactions (Figure 6ai). [75] The transparent, self-healing capabilities in aquatic conditions provide a route for unobtrusive underwater exploration (Figure 6a-ii). This ion gel electronics skin overcame the disadvantage of losing water due to the composition of amorphous polymer with chemically compatible ionic species. However, the stable mechanical and electrical interfaces between the hydrogel/gel with low Young's modulus and the relatively rigid printed circuit with high Young's modulus are hard to achieve due to the large difference in the elasticity modulus.
Material modification (such as surface super-hydrophobic modification) is also often used to prepare waterproof sensors. Niu et al. proposed waterproof electronic textiles by modifying silver nanoparticles (AgNPs)/cotton fabric (CF)/polydopamine (PDA) with 1H,1H,2H,2H-perfuorodecanethiol (PFDT), exhibiting a water contact angle of ≈152° (Figure 6b-i,ii)). [76] During the e-textiles were immersed in water for 50 h, their resistance keeps almost the same as the original value (Figure 6b-iii). However, this super-hydrophobic modification is prone to mechanical Figure 6. Design and fabrication of waterproof resistive sensors. a-i) Electronic skin with high stretchability and self-heal capability through highly reversible ion-dipole interactions. a-ii) Illustration of self-healed materials submerged in water for 3 h. Reproduced with permission. [75] Copyright 2019, Springer Nature. b-i) The fabrication process of the e-textiles. b-ii) Photographs of different liquid droplets on the surface of PFDT/Ag/PDA/CF e-textiles and CF. b-iii) Resistance stability of PFDT/Ag/PDA/CF e-textiles immersed in water for 50 h and the application of the interconnection with underwater. Reproduced with permission. [76] Copyright 2020, Tsinghua University Press and Springer. c-i) Photographs of hand gestures and their voltage profiles. c-ii) Photographs of human-machine interconnection with waterproof sensors. Reproduced under terms of the CC-BY license. [34] Copyright 2021, Wiley-VCH.
damage due to poor mechanical stability. The devices based on this method are easy to be damaged by friction or water impact.
Ultrathin interfacial encapsulation with stretchable elastomer is another effective method for waterproof sensors. But the materials and thickness of the encapsulation layer need to be elaborately designed and prepared. Common methods for fabricating the ultrathin encapsulation layer include spin-coating, thermal press, and spray coating. A robust waterproof LIG strain sensor was fabricated by LIG transfer and double encapsulation with silicone. [34] Real-time gesture recognition in both air and water was demonstrated. As shown in Figure 6c

Waterproof Sensor Systems
Most waterproof demonstrations reported so far are sensor-level and lab experiments that are difficult or impractical to extend to commercial applications. The commercial data gloves must be low-cost, lightweight, manufacturable in large quantity, and most importantly, waterproof at the system level. Such multiple demands have not been met. Wu's group explored an interfacial encapsulation strategy using a hydrophobic thin film to Figure 7. Waterproof data glove at the system level. a) Waterproof data glove consisting of a strain MXene fiber and ultralight-weight flexible printed circuit board (FPCB). b) The corresponding lookup table, and photographs of the finger-controlled underwater piano. c) Visualization of strain data via t-SNE dimensionality reduction and photographs of the application of remote control in the water. Reproduced with permission. [28] Copyright 2022, Elsevier.
develop waterproof data gloves. This waterproof data glove combined a strain MXene fiber and flexible printed circuit board (FPCB) through a layer-by-layer in-situ encapsulation technique (Figure 7a). The data glove acts as an underwater piano to play "Happy birthday to you" (Figure 7b). [28] Besides, assisted by a machine learning algorithm, the data glove achieved a 98.1% recognition accuracy for 20 kinds of hand gestures. It was then used to recognize underwater hand gestures and control the robot hand ( Figure 7c). However, in this work, mass production cannot be achieved because the fabrication technique is not compatible with the foundry.
Therefore, more efforts have been devoted to promoting the improvement in engineering, system integration, and commercialization of amphibious data gloves. For instance, Duan et al. proposed a waterproof, robust data glove equipped with gesture signal sensing, processing, and wireless transmission. Such data gloves were fabricated by directly integrating a stretchable strain sensor array and FPCB on commercial textile gloves based on Figure 8. Scalable waterproof data glove at the system level for amphibious interfaces and Metaverse interfaces. a) Data glove with low cost, lightweight, and high productivity. b) The resistive curve of the sensor integrated into the data glove with the strain between 60% and 100%. c) Excellent waterproofness of the sensor which could twist a bottle cup in both air and water. d) Modified neural network architecture, including environment-adaptive learning, and real-time inference for hand gestures. e) Remote controlling of guiding the vehicle to go along the "T" track. a-e) Reproduced with permission. [77] Copyright 2022, Tsinghua University Press and Springer. f) Remote controlling of a virtual reality shooting game. Reproduced with permission. [78] Copyright 2022, IEEE.
laser engraving and thermal transfer printing techniques. [77] This data glove featured low cost ($13.5 per unit; Commercial 5DT data glove: $1990), lightweight (25.5 g), and scalable for mass production (Figure 8a). The resistive sensors integrated into the data glove exhibited a high GF of 28.75 at the strain from 60% to 100% (Figure 8b). The data glove showed excellent waterproofness and was worn to twist a bottle cup in both air and water (Figure 8c). Through a modified neural network architecture, environment-adaptive learning and real-time inference for hand gestures were implemented, which achieved a nearly 100% recognition accuracy for 26 categories of hand gestures in both air and water (Figure 8d). The amphibious data glove was used to guide the vehicle to go along the "T" track in both air and water and also control a virtual reality (VR) shooting game (Figure 8e,f). [78] 4. Conclusion

Summary
In short, wearable resistive-type sensors are promising for gesture estimation due to their integration with ease, low cost, and easy signal acquisition. We here summarized three categories of resistive sensors for posture recognition based on the sensing mechanisms, including 1) monitoring the normal pressure at the www.advancedsciencenews.com www.advsensorres.com convex surface induced by jointing bending; 2) detecting the tensile strain at the convex surface of joints; 3) utilizing the compression strain at the concave surface of joints during bending. Besides, the typical sensors for each category are also outlined and discussed. The state-of-art resistive sensors have also been compared in a table based on the performance index (stretchability, cyclic stability, sensitivity, and response time) ( Table 2). Besides, common strategies for improving waterproofness including molecular design, material modification, and interfacial encapsulation are outlined. Such waterproof robust resistive sensorbased wearable systems for gesture recognition could revolutionize many applications including but not limited to human-Metaverse interfaces, human-machine collaboration, and augmented reality (AR).

Challenges
• Dynamic gesture recognition: Wearable resistive sensors could only characterize the bending angle during joint movement, which cannot achieve the spatial orientation of joints in space due to the lack of information about the spatial distribution of joints in space. Therefore, accurate dynamic gesture recognition cannot be achieved using resistive sensors alone. At present, algorithms based on machine vision are often used for the recognition of static hand gestures and the tracking of dynamic gestures. However, machine vision suffers from lighting conditions and object shielding. For example, in a situation where the light intensity is very high or the joints are blocked by other objects, the accuracy of posture recognition would be significantly reduced. However, machine vision could directly provide information about the spatial position of the joints by using visual information. Therefore, the multimodal fusion of visual image data provided by machine vision and somatosensory data provided by wearable resistive sensors could provide joint flexion angles and spatial position information of joints, which could provide a potential solution for accurate dynamic posture tracking.
As an example of static gesture recognition, a biologically inspired data fusion architecture that combines visual data with the somatosensory data acquired by SWCNT-based resistive strain sensors uses a convolutional neural network for vi-sual processing and then implements a sparse neural network for sensor data fusion and feature-level recognition, which could achieve high recognition accuracy under non-ideal conditions where images are noisy and under-or over-exposed. [17] The combination of visual information and somatosensory information may be a potential solution for precise dynamic gesture tracking that is not affected by environmental situations. • Interfacing with Metaverse: On the one hand, the interfacing technologies with Metaverse involve hardware and software components. The focus of the research is now on the hardware level (resistive sensors and their systems). There is still a gap between flexible wearable resistive sensor systems and Metaverse applications. Therefore, wearable resistive sensor systems cannot be used for Metaverse applications. In order to change this, more efforts should be made to create more imaginative demonstration applications. Meanwhile, the stability, cost, scalability, and sensing performance of wearable sensors should be well realized. Based on this, some commercial companies may develop upper-layer software frameworks that allow wearable sensor systems to access Metaverse applications. • Multimodal information acquisition: For more immersive interaction with Metaverse, it is not enough to rely on gesture recognition alone; multimodal information including gesture, posture, speech, eye movement, or even brain signals should be captured and fused in a multimodal fusion model. However, several key issues (i.e., tissue compatibility, safety, long-term stability, and cost) still need to be well addressed for brain-computer interfaces. [79] Furthermore, in next-generation intelligent human-Metaverse interfaces, a fused multimodal information model should be built. The model needs to collect all the information of a real-world human, process the multimodal information from wearable or implantable sensors, and create the final information model that could project all the information of a real-world human into his/her digital avatars in the Metaverse. In this approach, advanced multimodal fusion machine learning algorithms are required to decode mass multimodal information from different sensors. Edge computing devices should also be developed to reduce energy consumption and transmission latency of interfacing systems. [8]