Wearable integrated piezoelectric film sensor with 1 tensioning, bending, shearing and twisting detection 2 functions for human motion recognition

Human motion recognition using flexible/stretchable wearable sensors holds great 24 promise for human-machine interaction and biomedical engineering. However, to 25 measure the individual joint motion with multiple degrees of freedom, many 26 sensor networks are normally required and pinpointed onto the targeted area, 27 restricting body movement. This is due to the limitation of current wearable 28 sensors; inferring a sensor deformation based on the sensor's electrical signal is 29 challenging. A new concept of wearable sensor that can recognize how the sensor 30 deforms could radically solve this issue. Here, we report a wearable integrated 31 piezoelectric film sensor (i-PFS) capable of detecting basic deformations. To 32 achieve this, for the first time, we propose a novel design concept of using 33 uniaxially drawn piezoelectric poly L-lactic acid (PLLA) films to engineer 34 unimodal tension, bend, shear, and twist sensors that only respond to their 35 corresponding deformations with the enhanced piezoelectric response and self- 36 shielding function. Based on this, we construct the i-PFS by combining the four 37 unimodal sensors and demonstrate that the i-PFS can detect and differentiate 38 individual deformation modes, such as tensioning, bending, shearing, and twisting. 39 motions and subtle a virtual text- entry interface

sensors are required and pinpointed onto the desired positions, restricting the 58 naturalness of movements [20][21][22][23][24][25] . Fundamentally, such inefficiency and impracticality are 59 ascribed to the limitation of existing wearable sensors; gaining information on a sensor 60 deformation from the sensor's electrical signal is a huge challenge. For instance, the 61 wrist has multiple degrees of freedom; it can bend, twist, and even rotate 20 . We can 62 detect an electrical signal from a sensor attached to the wrist joint once deformed. 63 Conversely, it is not easy to infer the wrist deformation (i.e., the sensor deformation) 64 based on the detected electrical signal because any wrist motion generates a signal. It 65 is highly desirable to develop a new type of wearable sensor, which can provide 66 4 feedback on the sensor deformation, to radically solve the above problem. 67 Hence, we propose a novel concept of a wearable sensor comprising uniaxially drawn 68 piezoelectric poly L-lactic acid (PLLA) films, i.e., an integrated piezoelectric film 69 sensor (i-PFS), which can detect and differentiate four typical deformations, such as 70 tensioning, bending, shearing, and twisting.

71
Theoretical background 72 To develop the i-PFS, we first design four unimodal tension, bend, shear, and twist 73 sensors with the piezoelectric PLLA films, which only respond to tensioning, bending, 74 shearing, and twisting, respectively. The pristine piezoelectric PLLA film only presents 75 a shear piezoelectric coefficient (PC), i.e., d14, because of its helical structure 26 . It is 76 usually cut at a certain angle from the crystal orientation (i.e., drawing direction) of the 77 original PLLA film to respond to an external stimulus via modifying its PC 27 . We,   (Fig. 1d), which exactly agrees with our hypothesis.

123
Unimodal sensors fabrication and performance test 124 The piezoelectric property of uniaxially drawn PLLA film is strongly dependent on its 125 crystal orientation and crystallinity, which are dominated by fabrication conditions, 126 especially the drawing ratio (DR) 26,28 . We, therefore, prepared four uniaxially drawn 127 piezoelectric PLLA films with different DRs, such as 3.3, 3.7, 4.0, and 4.5 (Fig. 2a). As 128 shown in Fig. 2b, the crystal orientation improves with increasing the DR because a 129 Debye-Scherrer ring displayed at an initial DR of 3.3 gradually becomes three ellipses 130 at a maximum DR of 4.5, indicating the highly orientated α-crystal structure is formed 29 .

131
This behavior is observed by their melting thermograms as well. After a glass transition 132 7 temperature (Tg= 64°C), the cold crystallization temperature (Tc) of the drawn PLLA 133 films appears at different temperatures, and the lower DR, the higher Tc. This is because 134 the low chain orientation of the PLLA film requires more energy to form the ordered 135 arrangements and undergoes crystallization at a higher temperature (Fig. 2c). Besides, 136 there is a single melting peak (Tm) at around 170 o C, a typical α-crystal melting peak 26 , 137 but the β-crystal melting peak at Tm= 155 °C is not observed 30 , further confirming all 138 PLLA films are only compose of α-crystal. The crystallinity increases with increasing 139 the DR, and it reaches 60.9 % at a DR of 4.5 (Fig. 2d). As a result, the piezoelectric 140 PLLA film with the DR of 4.5 exhibits the most superior piezoelectric response among 141 them (Fig. 2e). Therefore, the PLLA film with the DR of 4.5 was selected and fabricated 142 into four unimodal tension, bend, twist, and shear sensors following the proposed 143 design concept (Fig. 2f). Notably, as the prepared PLLA film with α-crystal (i.e., 103    221 We then fabricate the i-PFS by stacking four unimodal sensors together, as shown in 222 Fig. 3f. The i-PFS was tensioned, bent, twisted, and sheared using the relevant machines. 223 The result proves that the i-PFS has enough capability to detect and differentiate 224 tensioning, bending, twisting, and shearing deformations, respectively ( Fig. 3f and 225 Supplementary Fig. 6). More importantly, this demonstrates that the four unimodal  Based on this, we designed a virtual text-entry interface system using the i-Glove in 260 conjunction with a convolutional neural network (CNN) algorithm for a finger-air-261 writing application, which detects finger movements with the i-PFS and transforms 262 them into corresponding characters (Fig. 4a,c and Supplementary Video 5). We chose 263 13 characters (i.e., "U", "O", "M", "0", "1", "2", "3", "4", "5", "6", "7", "8", "9") as 264 13 target classes. One participant was invited to create the data source for the finger-air-265 writing (see Method). As the character writing habit influences the output signal, the 266 writing style of each character was defined before collecting data to eliminate its effect.

267
For instance, we instructed the participant to write the letter "O" in clockwise and the 268 number "0" in anti-clockwise to distinguish them. Supplementary Fig. 9 displays the 269 output data of 13 characters of three trials, which shows good signal reproducibility. 270 We then adopted a LeNet-5 based CNN architecture consisting of two convolutional 271 layers for character classification (Supplementary Fig. 10). The classification result 272 shows that the accuracy using four channels can reach 89.7 % (Fig. 4d). We also   shear sensors are regarded as noises; these noises certainly increase with increasing the 351 tensioning intensity that is directly reflected by the tension sensor, as shown in Fig. 3b. Note that all parameters and applied forces used in these simulations (Fig. 1d) are the 426 same as the previous piezoelectric PLLA film simulations (Fig. 1a).  twisting, and shearing conditions. All conditions are the same as those used for 485 individual unimodal sensor (Fig. 3a).