A Wearable Flexible Tactile Sensor with Textile Microstructure for Wirelessly Recognizing Human Activity

Flexible tactile sensors have attracted special attention in the detection of human posture and activity. However, achieving multiple stimulus responses is particularly challenging in terms of structural design and has not been fully addressed. Herein, a novel structural design inspired by combining the characteristics of a strain sensor and e‐textile into a hybrid interface to obtain stretchable multiple stimulus‐responsive tactile sensors (stretching stimulus, pressing stimulus, and bending stimulus) is proposed. The sensitive layer of the sensor is designed as a textile microstructure. The sensor, which has a stable performance with excellent ohm property and durability (>2 years), possesses an extraordinary gauge factor (≈200) and a fast response time (≈2.48 ms) under tensile strain, a high sensitivity (≈2.35 kPa−1) in a low‐pressure range (≈9 kPa) under pressure. It has excellent performance for Human Activity Recognition by being attached to human skin. In addition, a Bluetooth device that tracks muscle activity in real‐time and transmits the output signals to a smartphone application is utilized. These results imply the utility of this sensor for a diverse application of robotic e‐skins or e‐muscles.


Introduction
Flexible electronics are of great importance in wearable electronic devices, [1][2][3][4] and have been used in flexible devices such as tactile sensors, nanogenerators, [5,6] photovoltaic devices, [7,8] actuators, [9] and neuroelectronic devices. [10][11][12] Especially, with DOI: 10.1002/adsr.202200051 the rapid development of humancomputer interaction, artificial intelligence, and wearable medical devices, flexible tactile sensors have become of great commercial value and research significance. [13][14][15][16][17][18][19][20][21][22] Various flexible tactile sensors, such as strain sensors, pressure sensors, temperature sensors, and humidity sensors have been used to emulate the touch-detection activities of human skin. The strain tactile sensor and pressure sensor are the most widely used due to their advantages such as strong tensile property, ultra-light weight, low cost, high sensitivity, and large range. [23][24][25] Studies of nanomaterials and micro/nano manufacturing have accelerated the development of flexible tactile sensors. [26,27] Flexible tactile sensors that use different types of sensing mechanisms such as piezoresistivity, [28][29][30][31][32] capacitance, [33][34][35][36] piezoelectricity, [37,38] and triboelectricity [39,40] have been explored. Tee [41] proposed a low-modulus self-healing foam material with embedded 3D electrodes to fabricate a flexible piezoresistive tactile sensor. Yang and Wei [42] demonstrated microstructured graphene electrode for ultrasensitive and tunable flexible capacitive pressure sensors. Hu and Yang [43] designed a piezoelectric tactile sensor array to sense and distinguish the size, position, and pattern of various external stimuli in real time. Lee and Guo [44] designed a triboelectric patch that can be used as an effective current collector, or attached to non-planar skin or clothing as a tactile sensor. However, structural design strategies for the recognition of tactile inputs with different stimuli reactivity have not been fully explored. For example, achieving accurate tensile-pressure-bending sensing in the dynamic conditions of human skin has been a problem for tactile sensors. Compared with conventional unstructured flexible strain tactile sensors, flexible tactile sensors with microstructured sensitive layer can achieve significantly higher sensitivity, faster response, and pressure sensing performance. Integration of multiple stimulus response functions in wearable electronic devices is an ongoing research stream, [45,46] but greatly complicates the tactile sensor system. Therefore, fabrication of flexible wearable sensors with simple structures and multiple tactile sensing functions remains a challenge.
Here, we report a flexible tactile sensor with textile microstructure sensitive layer to realize a response to multiple stimuli. We use a micrometer scale cross-woven textile microstructure template to manufacture the sensitive layer's structure on elastic polydimethylsiloxane (PDMS) substrate, embed carbon nanotubes (CNTs) in the textile microstructure area of the flexible substrate, add electrodes and encapsulate, piezoresistive flexible tactile sensor is prepared. The sensor responds to a variety of sensing information such as tension sensing, pressure sensing, and bending sensing, and can also recognize different bending directions. It is designed as a wearable smart patch attached to the human skin, collecting signals related to muscle activities and implements human activity recognition such as voice recognition and Morse code. In addition, connect it to Bluetooth device and realize wireless monitoring of sensor information in smart phone application. The development of this flexible tactile sensor enables the complete recognition of human activity signals and provides richer support for wearable devices, electronic skin, and soft robot applications.

Results and Discussion
Flexible tactile sensors are required in wearable electronic devices, electronic skin and soft robotics. The output signal of the wearable flexible tactile sensor can be transmitted wirelessly through Bluetooth device to a smartphone app when the flexible tactile sensor patches are attached to the human face or finger areas, respectively, for real-time activity recognition (Figure 1a). When the flexible tactile sensor responds to stretching stimulus, pressing stimulus or bending stimulus, its output signal changes due to the corresponding stimulus, and thereby achieves a recognition function (Figure 1b).
A sandwich structure of elastic polymer PDMS/CNTs/PDMS was designed to prepare a stable piezoresistive flexible tactile sensor, in which the bottom PDMS substrate has a textile microstructure (Figure 2), and the optical photograph of the PDMS substrate with textile microstructure are shown in Figure 2b. The sensor's underlying flexible PDMS layer provides mechanical support and the required flexibility, and in conjunction with the carbon nanotube layer provides strain resistance. The sensitive unit of the PDMS layer is treated with a surface microstructure, which strengthens the adhesion of CNTs to the PDMS surface. The CNTs separate or approach each other squeeze under different external stimuli, and the sensor's resistance changes in re-sponse to the reduction or increase in the number of conductive paths, and so enables detection and recognition of tactile information of different stimuli.
In this work, three kinds of flexible tactile sensors based on different sensitive unit structures were analyzed. The three sensitive unit structures are flat structure, particle structure and textile microstructure, respectively. And the surface scanning electron microscopy (SEM) morphologies of the three structures are shown in Figure S1 (Supporting Information). Package fixation ( Figure 2a) is performed to create a dense and stable CNTs network, which will provide consistent electrical and mechanical properties of the device. The top layer PDMS provides encapsulation to prevent physical damage and degradation. The flexible tactile sensor has a sandwich structure (Figure 2c,d), which is simple in structure and easy to operate, and the flexible device can tolerate bending. PDMS has flexibility, high stretchability, appropriate adhesion to human skin, and can be used as artificial skin.
The CNTs form uniformly distributed clusters, and the SEM image of CNTs is shown in Figure S2 (Supporting Information). The polyester fibers are woven to form a clothlike, flexible sensitive layer with textile microstructure, where the diameter of a single textile line is ≈16.4 μm (Figure 2e,f), and is critical for obtaining optimal electrical and mechanical properties of the device. This microstructure strengthens the interface adhesion of the flexible material PDMS and CNTs layer, and ensures that the electrical signal from the device changes differently under different mechanical deformations.

Electro-Mechanical Response of Flexible Tactile Sensor
Generally, human muscles experience stretching or other strains from negligible to 55%, or even higher, depending on the activity. However, human activities such as facial deformations of speaking or microexpression and finger flex activities are expected to occur within 20% in-plane strain. [47] The flexible sensor is clamped and monotonically stretched along the horizontal direction, which will demonstrate some behaviors that represent human life activities. Schematic of piezoresistive tactile sensor based on textile microstructure sensitive layer is shown in  The tactile sensor resistance change rate based on different structure sensitive layer under strain is shown in Figure 3b and Figure S4 (Supporting Information). In which, tensile strain is 100•ΔL/L 0 where ΔL is the change in length L compared to initial length L 0 of the sensor's sensitive unit, and relative resistance change is R′ = 100•ΔR/R 0 where ΔR is the change in resistance R compared to initial resistance R 0 . The largest gauge factor (GF) of the sensor with textile microstructure is 200 when tensile strain is more than 30%, which is more sensitive than flat structure and particle structure-based sensor. In which, the GF is the proportional change in resistance divided by the proportional change in length: When strain is less than 5%, the GF of textile microstructure based-sensor is about 146, but the GF of flat structure and particle structure both are less 146. When strain is more than 5% and less than 30%, the GF is ≈10 and when strain is more than 30%, the GF is about 200. However, GF of flat structure and particle structure based-sensors are less than 10% when the tensile strain is more than 5%. In a word, GF of textile microstructure-based sensor is higher than the two other types of sensors in any tensile range. The initial shape and size of the sensor do not affect the sensing result, because the tensile sensing performance of sensor is reflected by the relative change rate of resistance (ΔR/R 0 ) under the rela-tive strain (ΔL/L 0 ), that is, GF = (ΔR/R 0 )/ΔL/L 0 •100% to reflect the tensile strain sensing performance of the sensor, rather than the change of resistance (ΔR) to characterize the tensile strain performance. [48] Moreover, The response time of the sensor at 5% tensile strain is 2.48 ms, and the recovery time is 4.02 ms, which are compatible with a fast response ( Figure S5, Supporting Information).
Considering the micro-structure of sensor's sensitive unit, we also studied the pressure sensing performance of the sensor with different structure. The sensing performance of the sensors under pressure is shown in Figure 3c. The particle structurebased sensor hardly responds to pressure. And the other two kinds structure-based sensors have a certain response to pressure. However, the sensor with flat structure increases resistance with the increase of pressure, which tends to affect the tensile strain. The sensor with textile microstructure has a change of resistance decreasing with the increase of pressure, which is different from the response to tensile strain. In the pressure test, the sensor has a certain detection sensitivity S to the increase ΔP (kPa) in pressure: The maximum S of the sensor under pressure is 2.  In compression bending test, the change of sensor resistance is opposite to that of tensile bending, and the resistance decreases with the increase of compression bending degree, the R′ of the sensor decreases to about 62%. Therefore, the sensor also has a certain ability to detect and recognize different bending stimuli. Figure 3f shows R′ in resistance versus time plots of the textile microstructure-based tactile sensor under different tensile strain of 20%, 25%, 26%, 27%, and 28%. With the increase of tensile strain, R′ of the sensor also presents an increasing trend. R′ is also measured during stretching to = 5% at frequencies f S = 0.5, 0.1, 0.25, and 0.05 Hz (Figure 3g and Figure S7, Supporting Information). The peak R′ increased slightly as f S increased, but remained ≈500%. Figure 3h shows R′ versus time plot of our sensor under 500 cycles of strain at 5%, showing high repeatability and durability. Herein, we employed the first breaking point to evaluate the mechanical properties of the flexible sensor as shown in Figure 3i, the stress-strain curves of the sensor reflect the breaking stress of ≈15.81 N. In addition, the flexible tactile sensor with textile microstructure was prepared two years ago. After multiple external stimulation tests, it was stored in room temperature environment for 2 years. The IV curve is shown in Figure S8 (Supporting Information), which still maintains good ohmic characteristics.

Micro-Basic Strain Analysis of Flexible Tactile Sensor
To understand the main factors of sensor sensing under different stimulus response behaviors (Figure 3), the strain analysis of the flexible tactile sensor with textile microstructure (Figure  4) was carried out respectively under the initial state, stretching state, pressing state and different bending states (tensile bending and compression bending). The total resistance of sensitive unit based on CNTs/PDMS is a function of both the resistance of each conducting CNTs particle and of the polymer matrix PDMS. The charge carriers in a flexible device move between two localized states by jumping or tunneling. [49][50][51][52] Therefore, the modulation of tunneling distance for a stretchable sensor under strain causes major changes in resistance by modifying the conducting channels.
During a uniform stretching test, the CNTs physically disconnect from each other during the stretching process with strain increasing, increasing the path of electron transmission and the number of conduction paths decreases (Figure 4a, blue circles), and R of the sensor increases. For a uniform pressing test, the change in the conductive path is the opposite of the stretching process. the CNTs physically get together from each other during the pressing process with pressure increasing, decreasing the path of electron transmission and the number of conduction paths increases (Figure 4a, red circles), and R of the sensor decreases.
However, bending deformation can cause complex changes in the electrical information of the device. In this experiment, the flexible tactile sensor was attached to a curved carrier composed of polyethylene terephthalate (PET) film with a thickness of 0.05 mm. Tensile bending and compression bendingdeformation tests are performed on the sensor (Figure 4b). In the bending tests, we define that the bottom of the sensor is attached to the curved carrier PET, and CNTs in sensitive layer are divided into an upper part and a lower part. When the sensor is bent downward, the lower part of the sensor is almost unchanged due to the negligible strain caused by the contact with the bending carrier. However, CNTs in the upper part away from the curved carrier is changed from the contact state to the disconnection state, resulting in a decrease in the conductive path and an increase in resistance because of the tensile bending strain (Figure 4b, blue circles). On the contrary, when the sensor is bent up, the bending of the carrier produced compression bending to CNTs, as shown in the red circles in Figure 4b, resulting in a large amount of aggregation. The resistance of the sensor is reduced. These multiple stimulus-responsive sensors (including tensile, pressure, or bending sensing) are more conducive to wearable applications such as voice recognition, Human Motion Recognition, Heart Rate measurement, Brain-computer integration, and soft robotics.

Application Demonstration
After confirming the mechanical reliability, recognition function of different tactile stimuli, and high sensitivity of the sensor developed in this work, we proposed a wearable application of the flexible tactile sensor patch attached to the human finger joints and face.
Attach the flexible sensor to the back of the finger joint, as shown in Figure 5a and Figure S9 (Supporting Information), to test the degree of finger curvature in real time. Figure 5b shows the function diagram of the obtained signal over time. Figure S10 (Supporting Information) shows the relationship between the obtained signal and the degree of finger curvature. It can be seen that R′ also increased with bending angles in a sensor that had been applied to a human index finger. At the maximum bending of the joint, R′ is 3000%. Further, we use flexible tactile sensors in Morse code applications. And the bending for a moment is defined as the dot signal in Morse code, the bending for a period of time is defined as the dash signal in Morse code. We carried out the Morse code application of the SOS signal as shown in Figure 5c. The sensor resistance change could accurately represent the "SOS" Morse code signal.
In addition, the flexible sensor is attached to the human face to recognize the expansion and relaxation of the human face during pronunciation of the names of the five vowels "AEIOU" (Figure 5d,e and Figure S11, Supporting Information). When pronouncing "A," the face expands first (tensile bending), and the resistance of the flexible sensor increases with this expansion. When the pronunciation of "A" ends, the resistance decreases as the face relaxes and recovers. At this time, due to a small interference caused by facial movement, the sensor resistance value is balanced at a new value. Then, during pronunciation of "E," the flexible sensor produces a new trend of increasing resistance with the expansion of the face, then decreasing resistance after the sound ended. During pronunciation of "I," the face expanded and recovered to a greater extent than for "A" and "E," so the flexible device will have a correspondingly greater trend in resistance increase and recovery decrease. During pronunciation of "O," the flexible sensor flexes inward (compression bending) with the facial muscles, so the change in resistance decreases. During pronunciation of "U," the face produces a small concave deformation, similar to the changes during pronunciation of "O," and the flexible sensor produces a small compression bending action with the facial depression, so the resistance shows a weak decreasing trend.
The sensor could also recognize the pronunciations of the words. Here, we did two pronunciation tests for words such as "Pressure Sensor" (Figure 5f) and "Nankai" (Figure 5g). In the pronunciation tests, the flexible sensor produces stretching and tensile bending deformations of the attachment carrier as the facial vowel is pronounced. The experimental results show that the sensors have the same change trend during multiple pronunciation recognition of the same word.
The flexible tactile sensor could also recognize different gestures (including tensile bending and compression bending). We connect the sensor to a Bluetooth device to generate a wireless signal, and use a mobile phone APP to monitor the sensor's performance in real time (Figure 6a). Here, the tactile sensor is attached to the backside of the index finger joint (Figure 6b-i) and the inside of the joint (Figure 6b-ii) to quantify the response to the same finger bending motion. The Bluetooth system is used to track finger bending activity in real time, and to wirelessly transmit the output signal to the APP. When the sensor is attached to the backside of the finger joint, the tensile bending strain is generated when the finger is bent, and the resistance of the sensor increases with the bending of the finger (Figure 6d). How-ever, when the sensor is attached to the inside of the finger joint, the compression bending strain is generated when the finger is bent, the resistance of the sensor decreases with the bending of the finger (Figure 6e). In the display of the smartphone device, the signal changes as the finger bends (Movie S1, Supporting Information).
The sensor could also recognize different guessing-game gestures (such as "paper," "rock," "scissors") ( Figure 6c). The resistance of the sensor responds differently to bending, depending on whether it is tensile bending deformation or compression bending deformation when attached to the backside of the finger joint. When attached to the backside of the finger, the sensor could recognize "paper," "rock," "scissors." As shown in Figure 6f, the "paper" gesture is the initial state of the sensor, and indicates that five fingers are normally extended (Figure 6c-i,iv). During the "Rock" gesture, all five fingers are all bent (Figure 6cii,v), and the sensor resistance increases with the deformation of tensile bending. On the contrary, during the "Scissors" gesture, the index finger and middle finger bend slightly in opposite direction of the "Rock" gesture ( Figure 6c-iii,vi), and then the sensor produces a compression bending deformation, so the sensor resistance decreases with the deformation of compression bending. These results prove that the developed flexible sensor can recognize the complex tactile stimuli and has high mechanical reliability. The device can be used to recognize a variety of human life activities, and could have wide applications in future wearable devices.

Conclusion
In summary, we have devised a stretchable multiple stimulusresponsive tactile sensor based on textile microstructure sensitive layer. This sensor has good tensile sensing, pressure sensing and bending sensing properties. It is also used to recognize human activity, such as voice recognition (vowel recognition) and gesture recognition (finger-guessing game). In addition, connect it to a Bluetooth device and realize wireless monitoring of sensor information in smart phone application. Therefore, it is an excellent candidate for wearable and portable application. The fine differentiation of detailed movements can be further investigated for e-skin applications, such as expression recognition, brain-machine interface, and complex robotics applications.

Experimental Section
Fabrication of Flexible Substrate with Textile Microstructure: The warp and weft yarns were interwoven in a ratio of about 10:18 to form a textile fabric, and a polypropylene adhesive with a thickness of 20 mm was pasted on one side of the textile fabric to become a textile tape. The textile tape was cut to form a rectangle with a size of 3 mm × 20 mm, then stuck in the middle of a square petri dish (10 cm × 10 cm). The PDMS (Sylgard 184, Dow Corning, USA) precursor and curing agent were prepared in a mass ratio of 10:1 and stirred evenly, then 10 g of the PDMS solution was evenly dropped into a petri dish with a textile microstructure rectangular template, exposed to vacuum for 10 min, then cured at 80°C for 2 h. Then the cured flexible PDMS membrane was peeled from the petri dish, to obtain a flexible substrate with a textile microstructure (Figure 2a,b).
Fabrication of Flexible Tactile Sensor: Firstly, PDMS flexible substrate with rectangular textile microstructure grooves was prepared. Secondly, two conductive tapes as electrodes were attached to two narrow ends of the textile microstructure grooves, and 500 μL of carbon nanotube solution was dropped onto the flexible substrate, which was then annealed at 70°C for 30 min to yield a microstructured conductive film was obtained, among them, the CNTs solution was purchased from Beijing Boyu Gaoke New Material Technology Co., LTD., with an outer diameter of 30-50 nm and a length of 10-20 μm. It was dispersed in deionized water according to the mass ratio of 1:10 and dispersed into relatively uniform CNTs cluster by ultrasonic. Finally, 1 mL of the prepared PDMS solution was dropped over the sensitive cell and cured at 80°C for 2 h to encapsulate the flexible device. After curing PDMS, a flexible tactile sensor with textile microstructure was prepared (Figure 2a).
Characterizations: SEM characterizations of CNTs nanomaterial and the textile microstructure of flexible substrate morphology were performed using a field-emission microscope (Apero S, Thermo Scientific). All electrical measurements were performed using a Source-Meter measuring instrument analyzer (Keithley 2400), and the equipment was tested at room temperature in ambient air.
Tactile Testing of Sensor Devices (Tension, Pressure, Bending): Tensile and bending strain tests used a flexible electronic tester to uniformly apply different actions to the flexible tactile sensor in different initial states and moving directions, and used a Keithley 2400 SourceMeter to record resistances. The pressure strain test used a MARK10 ESM303 electric tensile test bench to apply a uniform displacement at a constant speed of 30 mm min −1 to apply pressures of different magnitudes to the sensor, and also used a Keithley 2400 Source Meter to record resistances.
Recognizing Human Facial Muscle and Finger Activities: To recognize the expansion or relaxation of human muscles in real time during life activities, the sensing devices were attached to the face or finger of a human subject, and the two ends of the sensor electrodes were connected to the Keithley 2400 SourceMeter. When the human face makes different vowel sounds, the facial muscles deform to varying degrees and the deformations affected the electrical signals through the sensor. Similarly, when the hand performed different actions during "paper, rock, scissors," distinct electrical signal changes were generated and used to distinguish among the actions. All human subjects involved in the human activity signal tests provided informed consent, and the study protocol (no.NKUIRB2023004) was approved by the Institutional Review Board officeat Nankai University. The skin shown in the figures and video are those of Y.D., who has given her consent to publish these images and movies.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.