System-Level Integration and Multimodal Signal Acquisition in Electronic Skin: A Review and Demonstration

E-skin represents a new generation of soft electronic systems capable of mimicking the sensory and functional properties of human skin. By detecting mechanical and thermal stimuli such as pressure, strain, temperature, and slip, E-skin enables machines to perceive their surroundings in a human-like manner. Since the early work by Someya and colleagues that introduced flexible skin-inspired electronics, research in this field has advanced rapidly across materials science, device engineering, and system integration^1,2.

Early developments in E-skin primarily focused on creating stretchable and conductive materials that could retain electrical stability under mechanical deformation. Conductive polymer composites, carbon nanomaterials, and liquid-metal-based electrodes have been widely explored to improve flexibility and sensitivity^3,4,5. These efforts established the foundation for reliable tactile sensing. However, as research expanded toward robotics and wearable electronics, the bottleneck shifted from materials to system-level integration, where challenges such as large-scale signal acquisition, low-noise amplification, and real-time data processing became central concerns^6,7.

Recent progress highlights the growing importance of multimodal tactile sensing and embedded signal acquisition, which have transformed E-skin from a passive sensing interface into an active perception platform. Jeon et al.⁸introduced one of the earliest multimodal tactile systems integrating pressure, temperature, and slip sensing, while Lee et al.⁹demonstrated real-time tactile perception by embedding signal processing circuits directly into the sensor array. Similarly, Wang et al.¹⁰proposed a bioinspired robotic tactile system capable of recognizing fluid interaction forces with high temporal resolution. These studies reveal a clear shift toward integrated sensing architectures capable of complex perception and classification.

Beyond robotics, E-skin has also shown broad potential in health monitoring and biomedical interfaces. Continuous measurement of physiological signals-such as pulse, temperature, and electromyography-enables personalized healthcare and long-term rehabilitation^11,12. Wu et al.⁷ summarized recent progress in multimodal wearable sensors, emphasizing challenges in signal decoupling and stability during prolonged operation. Yet, as AlShaibani et al.¹³ pointed out, the field still lacks standardized evaluation protocols, making it difficult to compare sensing accuracy among systems. Establishing unified test frameworks is therefore essential for future clinical and robotic adoption¹³.

Given these developments, this review focuses on system-level architecture and signal acquisition strategies for E-skin. It emphasizes how multimodal sensing, circuit design, and data fusion collectively determine system performance and functional scalability^14,15. To demonstrate these principles, a 36-channel hybrid-frequency tactile sensing platform developed by the authors is introduced as an example of real-time multimodal signal fusion and robotic interaction. This demonstration serves as a bridge between conceptual design and practical robotic skin applications.

Overall, this review provides a comprehensive system-level perspective on the development of E-skin, bridging the gap from multimodal sensing principles and hardware architecture to intelligent perception. By emphasizing signal acquisition strategies, data fusion, and practical robotic implementation, it highlights both the recent progress achieved and the key challenges that remain in realizing adaptive, scalable, and reliable E-skin systems.

Overview of electronic skin architectures

E-skin can be regarded as a hierarchical soft electronic system that integrates sensing, acquisition, processing, and communication layers into a unified structure. Each layer plays a distinct role in transforming external physical stimuli into meaningful digital information that can be interpreted by machines or humans. Figure 1 provides a conceptual schematic of this four-layer architecture and the corresponding signal flow.

At the sensing layer , flexible and stretchable sensor arrays detect mechanical or thermal stimuli such as pressure, strain, or temperature. These elements are fabricated on elastomeric substrates and often employ microstructured geometries to improve sensitivity and response linearity. Conductive polymers and nanocomposites provide high compliance and stable electrical performance under large deformation^3,5. In multimodal E-skin systems, multiple physical stimuli-such as pressure, strain, and temperature-can be detected either by integrating diverse sensing units or by employing microstructured device architectures that decouple these stimuli within a single sensor⁴.

The acquisition layer serves as the electrical bridge between sensors and electronics. Its main functions include amplification, filtering, and analog-to-digital conversion. In modern E-skin, flexible printed circuits or thin-film transistor matrices are widely used to realize active-matrix addressing and multiplexed readout. These approaches reduce wiring complexity while maintaining high spatial resolution and low power consumption. The design of this layer determines the system's noise floor, sampling rate, and overall data fidelity.

Next, the data processing layer manages real-time interpretation of tactile information. Embedded microcontrollers or field-programmable gate array (FPGA)-based processors execute local filtering, feature extraction, and preliminary classification of tactile events. This distributed computing architecture minimizes latency and enables adaptive feedback control. Integration with wireless or edge-AI modules further supports autonomous robotic operation and wearable applications.

Finally, the interface or application layer links the E-skin platform to external systems such as robotic manipulators, prosthetic limbs, or healthcare monitors. Processed tactile data are converted into control commands, visual outputs, or diagnostic indicators. For robotics, this layer provides grasp-force feedback and texture recognition; for biomedical use, it supports continuous physiological monitoring and remote data transmission.

Together, these four layers form an end-to-end pathway that converts mechanical or thermal stimuli into actionable information. The hierarchical and modular design allows each functional block to be optimized independently while maintaining overall mechanical flexibility. This overview establishes the structural framework for the following sections, where the sensing mechanisms, signal acquisition methods, and intelligent perception strategies of E-skin systems will be analyzed in detail.

Sensing mechanisms and transduction principles

E-skin converts mechanical and thermal stimuli into electrical signals through several transduction mechanisms. The most widely used are piezoresistive , capacitive, self-powered piezoelectric, and optical mechanisms (Figure 2). Each approach offers distinct advantages in sensitivity, dynamic range, and energy consumption.

Piezoresistive and capacitive sensing: Piezoresistive sensors (Figure 2A) change their resistance when deformation alters the conductive network of the sensing film. They are simple to fabricate, highly compatible with flexible substrates, and provide a broad measurement range. Typical materials include carbon-based composites and conductive polymers. Capacitive sensors, in contrast, rely on changes in capacitance C=εA/d caused by the mechanical deformation of a dielectric layer. They feature high stability and low noise and are well-suited for tactile imaging. Jeon et al.⁸ demonstrated a multimodal E-skin combining piezoresistive and capacitive elements to sense pressure and temperature simultaneously with minimal interference.

Self-powered mechanisms: Triboelectric (Figure 2B) and piezoelectric (Figure 2C) transducers generate electrical signals directly from mechanical motion. Piezoelectric films produce charge separation under stress, while triboelectric layers generate potential differences through contact electrification. These self-powered mechanisms are effective for dynamic stimuli such as vibration and slip. Wang et al.¹⁰ developed a robotic tactile system based on a triboelectric interface capable of perceiving fluid interactions with high temporal resolution. Although less suitable for static pressures because of charge leakage, combining them with piezoresistive or capacitive layers enables complementary multimodal sensing.

Optical and other emerging mechanisms: Optical sensing (Figure 2D) converts mechanical deformation into changes in light intensity or interference within waveguides or microcavities. It offers ultrahigh sensitivity and immunity to electromagnetic noise, but requires precise alignment¹⁶. Other approaches-including magnetic¹⁷, thermoelectric¹⁸, and impedance-based methods¹⁹-are also being explored to broaden E-skin functionality.

Toward multimodal integration: Modern E-skin systems increasingly combine multiple mechanisms to detect and distinguish various stimuli. Lee et al.⁹ achieved real-time pressure, temperature, and slip detection through integrated sensor arrays and embedded processing circuits. Such hybrid architectures advance E-skin from passive sensing to active perception, laying the foundation for the system-level integration discussed in the next section.

System integration and signal acquisition

The performance of an E-skin system depends not only on the properties of its sensing elements but also on how the signals from those sensors are acquired, processed, and transmitted. System-level integration links the sensing layer with embedded electronics, forming a unified tactile perception platform that is capable of real-time responses. Figure 3 illustrates a representative system architecture that includes sensor arrays, readout electronics, data acquisition, and communication modules.

Architecture and hardware design: Modern E-skin architectures are typically organized in a modular structure consisting of three hardware subsystems: the sensor array, the readout and multiplexing circuits, and the embedded controller or processing unit. This hierarchical configuration balances spatial resolution with system scalability. In most implementations, large-area sensor arrays are arranged in a matrix format, where each node corresponds to an individual sensing element. To reduce wiring complexity and crosstalk, time-division multiplexing (TDM) or row-column scanning strategies are employed. The readout circuit includes impedance matching, low-noise amplification, filtering, and analog-to-digital conversion (ADC). Takei et al.¹⁴ emphasized that the integration of sensing materials with low-power CMOS readout chips is crucial for realizing compact and efficient E-skin platforms. Embedded microcontrollers (MCUs) or FPGAs then manage channel selection, data buffering, and preprocessing. To maintain flexibility, rigid silicon chips are often mounted on soft interposers or flexible printed circuit boards (FPCs). Wireless communication modules-such as Bluetooth Low Energy (BLE) or Wi-Fi-are frequently integrated for wearable applications, whereas robotic platforms often adopt wired serial communication for real-time control.

Signal acquisition strategies: Signal acquisition in E-skin systems must handle high channel density , low signal amplitude , and wide dynamic range . The tactile signals from each sensor are typically in the millivolt to microvolt range, requiring high-gain differential amplifiers and noise suppression circuits. Common approaches include correlated double sampling (CDS), chopper-stabilized amplifiers^20,21, and synchronous detection methods²² to minimize drift and 1/f noise. Viteckova et al.²³ analyzed data acquisition frameworks for wearable tactile systems and emphasized the trade-off between sampling speed and energy efficiency. Multi-channel ADCs with sequential scanning are widely used, but they introduce a sampling delay across the array. Parallel sampling, while faster, increases power and data throughput requirements. Thus, hybrid solutions that combine sequential and parallel acquisition have gained interest for scalable systems²⁴.

Data fusion and intelligent perception: After acquisition, tactile signals from E-skin systems must be processed to extract meaningful features and support intelligent perception. Signal processing serves as the bridge between raw sensor output and physical understanding, enabling noise suppression, feature extraction, and multimodal data fusion (Figure 4).

Pre-processing and feature extraction: Raw signals are often distorted by drift, offsets, or electromagnetic interference. Baseline correction, adaptive or low-pass filtering, and differential amplification are commonly employed to improve signal-to-noise ratio. Subsequent feature extraction transforms time-domain data into compact descriptors-such as amplitude, duration, and rate of change-while spatial filters reconstruct tactile maps across sensor arrays. Frequency-domain analysis using Fourier or wavelet transforms further reveals vibration and texture information.

Multimodal fusion and interpretation: In multimodal E-skin systems, signals from pressure, strain, temperature, or vibration sensors are combined to form unified tactile representations. Data-level and feature-level fusion improve robustness, while decision-level fusion integrates the results of multiple classifiers. Recent studies have shown that machine-learning algorithms-including principal component analysis and convolutional neural networks-can enhance tactile event recognition²⁵. Furthermore, multimodal fusion frameworks combining tactile and proprioceptive data have been explored to improve classification accuracy²⁶.

Toward intelligent perception: At higher abstraction levels, E-skin data are interpreted to infer object properties, contact states, and interaction intent. Pattern-recognition and regression models classify tactile events such as grasp, slip, or texture, whereas clustering and spatiotemporal analysis reveal hidden relationships among distributed signals²⁷. The integration of artificial-intelligence-based reasoning enables context-aware perception and real-time feedback for robotics and wearable healthcare.

In summary, signal processing and data interpretation convert raw tactile information into actionable insight. Advances in adaptive filtering, multimodal fusion, and data-driven modeling are propelling E-skin systems toward higher perceptual accuracy and autonomy.

Applications and demonstrations

Flexible tactile systems have found applications across diverse domains, ranging from robotic manipulation to wearable healthcare and human-machine interaction. These technologies demonstrate how multimodal sensing, compliant materials, and embedded electronics can converge to form intelligent interfaces capable of real-time feedback and adaptive control (Figure 5).

Robotic manipulation and grasping: In the field of robotics, flexible tactile systems provide essential feedback for safe and dexterous manipulation. By integrating capacitive, thermal, and vibrational sensing, these systems enable adaptive grasping and real-time force modulation when handling delicate objects. Such tactile feedback also facilitates edge detection, pose estimation, and compliance control, significantly enhancing human-robot collaboration^28.

Wearable and healthcare monitoring: For wearable health applications, flexible sensor patches enable continuous physiological monitoring during daily activity. Recent progress in flexible and wearable biosensors has greatly expanded the scope of continuous health monitoring. Song et al.²⁹summarized advances in soft, skin-conformal sensing platforms capable of real-time detection of physiological signals from body fluids and vital signs. Such biosensors combine high flexibility, low power consumption, and biocompatibility, offering reliable long-term operation for applications in personalized medicine and preventive healthcare. Similar systems have been explored for gait analysis, muscle activation monitoring, and rehabilitation feedback^30,31.

Human-machine interaction and assistive systems: Soft, skin-inspired tactile electronics are increasingly employed to enable intuitive communication between humans and machines. By translating mechanical or physiological signals into interpretable electrical responses, these flexible systems provide natural interfaces for robotic control and assistive technologies. Liu et al.³² developed a wireless electronic-skin platform that functions as a bidirectional human-machine interface, allowing gesture-based control of robotic devices and feedback to the user through integrated sensors and actuators. Complementary advances in wearable e-skin for health monitoring have expanded its assistive role for patients and the elderly. Sun et al.³³ summarized how conform-able biosensing networks and adaptive data interpretation contribute to continuous health support and rehabilitation feedback. Together, these studies demonstrate how tactile electronics enhance perception, communication, and assistance in next-generation interactive systems.

Demonstration using an integrated 36-channel tactile acquisition platform: To illustrate practical system integration, a 36-channel tactile acquisition platform developed by the authors is presented as a representative demonstration³⁴. The platform integrates a flexible sensing array with a custom acquisition board containing low-noise amplifiers, analog multiplexers, and a 14-bit ADC. A microcontroller manages data synchronization and communication via USB or Bluetooth Low Energy, while an on-board CPLD controls timing and buffering. Instead of a single fixed rate, the system performs three synchronous sampling streams-2 kHz for acoustic signals, 1 kHz for acceleration and tactile signals, and 100 Hz for temperature and light-so that each sensing modality is sampled at an appropriate rate, improving bandwidth efficiency without sacrificing perceptual performance. Real-time data are visualized and analyzed on a host computer, demonstrating scalable tactile perception suitable for both robotic and wearable platforms.

In summary, the four application domains demonstrate how flexible tactile electronics are transforming sensing into perception. From robotic manipulation to wearable health and interactive systems, the integration of multimodal sensors, adaptive acquisition, and intelligent processing enables responsive and efficient tactile interfaces, paving the way for autonomous and human-adaptive e-skin platforms.

Despite rapid progress in materials, system integration, and data interpretation, electronic skin (E-skin) technologies still face several barriers before large-scale practical deployment can be achieved. Improvements are needed in mechanical reliability, scalability, and intelligent autonomy to realize truly perceptive electronic platforms. E-skin depends on flexible and stretchable materials that can sustain sensitivity under continuous deformation; however, long-term mechanical fatigue, environmental degradation, and baseline drift remain unresolved. Achieving uniformity and repeatability in large-area fabrication also presents major obstacles. Future research should therefore focus on scalable, low-cost micro/nanomanufacturing and encapsulation methods that ensure durability and biocompatibility. Integrating multiple sensing modalities-pressure, strain, temperature, vibration, and acoustic signals-within compact hardware inevitably introduces wiring complexity, crosstalk, and energy consumption. Efficient multiplexing, low-noise analog front ends, and embedded power management will be essential to improve autonomy. Self-powered sensors, energy-harvesting modules, and system-on-chip architectures are promising routes toward fully untethered operation. Meanwhile, the massive and heterogeneous data generated by E-skin requires robust and explainable interpretation. Machine-learning-based algorithms have enabled event classification and tactile perception, yet real-time, low-latency decision-making under limited power and bandwidth remains challenging. The incorporation of neuromorphic and edge-AI concepts could enable local processing with minimal data transmission, supporting context-aware and adaptive behavior. Looking forward, next-generation E-skin systems will likely integrate self-healing materials, hybrid energy systems, and intelligent acquisition circuits to achieve long-term, autonomous operation. Standardized communication protocols, data security, and human-safe biocompatibility will be critical for clinical and collaborative robotic applications. As flexible electronics continue to merge with artificial intelligence, E-skin is expected to evolve from passive sensing layers into active, perceptive systems capable of understanding and responding to complex human-environment interactions.

Figure 1: Conceptual schematic of four-layer architecture. An electronic skin (E-skin) system consists of layered components that collectively enable tactile perception and interaction. The sensing layer detects mechanical and thermal stimuli using flexible, microstructured sensors; the acquisition layer amplifies and digitizes these signals; the processing layer interprets tactile data in real-time; and the interface layer converts this information into robotic or biomedical feedback. Please click here to view a larger version of this figure.

Figure 2: Different types of E-skin. (A) Piezoresistive sensors detect deformation-induced resistance changes within conductive films. They are easy to fabricate, compatible with flexible substrates, and offer a wide sensing range using carbon composites or conductive polymers. (B) Triboelectric transducers generate electrical signals from contact electrification, providing self-powered detection of dynamic stimuli such as vibration and slip with high temporal resolution. (C) Piezoelectric sensors convert mechanical stress into electrical charge, enabling self-powered sensing of dynamic forces and complementing other mechanisms for multimodal E-skin applications. (D) Optical sensors translate mechanical deformation into variations in light intensity or interference within waveguides, offering ultrahigh sensitivity and strong resistance to electromagnetic interference. Please click here to view a larger version of this figure.

Figure 3: Electronic skin architecture schematic. This schematic delineates the integrated architecture and signal processing-transmission workflow of electronic skin (e-skin), encompassing five core functional layers. Initially, external physical stimuli (mechanical or thermal) are transduced into raw electrical signals by the sensing layer, which leverages piezoresistive, capacitive, piezoelectric, or optical sensing mechanisms. These raw signals undergo amplification, filtering, and preliminary conditioning in the acquisition layer equipped with integrated amplifiers and filter modules. Subsequently, the processing layer performs analog-to-digital conversion (ADC), feature extraction, and information interpretation via flexible processors or microcontroller units (MCUs). Following this, the communication layer enables wireless/wired data transmission through technologies such as Bluetooth, NFC, or flexible transmission links. Finally, the terminal reception layer facilitates functional execution or information reading by machines or human users, thus establishing a comprehensive signal chain for E-skin-driven intelligent sensing and interactive applications. Please click here to view a larger version of this figure.

Figure 4: Pressure schematic of E-skin. This electronic skin pressure schematic visualizes the global force distribution by compiling data from all sensor channels. The pressure on each strain sensor, which corresponds to its recorded voltage, is indicated by a color gradient; greater pressure/voltage is denoted by a darker color. This display scheme facilitates intuitive observation of the pressure profile across the e-skin. Please click here to view a larger version of this figure.

Figure 5: Flexible tactile system. The flexible tactile system is composed of a sensor board and a data acquisition board, both fabricated on flexible printed circuit (FPC) substrates. The small black squares are viscoelastic polymer sensing elements, which exhibit both elastic deformation and viscous flow under mechanical load. The acquisition board collects data from all 36 sensor channels on the sensor board and transmits information wirelessly via WIFI. Please click here to view a larger version of this figure.