**1. Introduction**

The post stroke gait disturbance is one of the major complications that requires a long-term rehabilitation and limits the patient's activities of daily living [1]. The improved gait function is an important factor in returning to social life and thus is an essential goal of rehabilitation therapy [2]. While 60% to 80% of stroke patients regain their independent ambulation function, many exhibit hemiparetic gait pattern for the rest of their lives due to unilateral neuromuscular weakness leading to gait asymmetry [2–4]. The critical consequences of the impaired walking ability after a stroke include a reduction in gait speed, shorter step and stride lengths, and an increased fall risk [5,6]. These residual deficits are mainly caused by muscle weakness and imbalance, decreased weight support on the affected side, and asymmetrical intralimb coordination [7,8]. For the patients with these deviations to regain walking abilities, clinical treatments commonly rely on traditional rehabilitation approaches such as neuromuscular re-education, lower limb strength training, and balance training for weight shifting and gait pattern training. While the patients and healthcare providers strive to seek for more effective gait therapy methods, there are not many long-term remedies or set devices for facilitating these treatment options in clinical settings.

With fast growing advances in technology, various forms of wearable measuring-recording devices and sensors have been developed for the health care system. For gait rehabilitation, force sensitive resistor sensors (FSR) are one of the most commonly used sensor types for analyzing gait pattern or plantar pressure, but is known to have the disadvantage of being deformed in time as a response to repetitive force applied. Another drawback of the FSR sensor is the need to be constantly calibrated (good for approximately 100 uses) and its inability to distinguish between load changes in respect to the weight bearing level difference generated during walking [9,10]. Another popular type considered for gait biofeedback sensors is piezoelectric material, which has a high impedance, low noise, and susceptibility to electrical interference. However, its weakness is having a limited pressure range and non-linearity in repeated signal output [9]. According to a study by Aqueveque et al. (2018), capacitive sensors can be used for developing an in-shoe device that measures plantar pressure [9]. A capacitive sensor basically consists of two conductive plates that are separated by dielectric material. Modifying the distance between the conductive plates generates a variation of the capacitance and this change in capacitive value can be interpreted as plantar pressure change [9]. This study developed a capacitive insole sensor to analyze plantar pressure and with the aim to provide real-time feedback on the patients' gait rehabilitation processes.

The assessment of gait rehabilitation is usually made using clinical scales, but many of the neuromuscular disabilities are still being evaluated manually with analog scales [11–15]. Thus, it is difficult to verify the accuracy in measurement methods, and both the intra- and inter-personal evaluation results have low reliability. A practical device that can quantitatively analyze the characteristics of hemiparetic gait pre and post rehabilitation is substantially needed. A recent study by Ngueleu et al. (2019) stated that, although using pressure-sensing insoles for identification of the step count is a promising approach in gait analysis, the accuracy of the activity monitors in step counting remains limited and that there appears to be no consensus for optimal positioning and the number of sensors for insoles [10].

In stroke rehabilitation, the basic gait analysis parameters include the plantar pressure distribution, the step count, the stride time, velocity (or gait speed), the center of pressure (CoP), the coefficient of variation (CV) and the phase coordination index (PCI) [16–23]. Among these indices, plantar pressure and PCI have shown to be relatively more sensitive measures in analyzing the bilateral coordination or asymmetry of locomotion and balance, which are meaningful variables [16,17,24,25]. According to previous studies, the wearable gait analysis system [26], such as with insole pressure sensors, can analyze gait parameters including gait velocity, cadence, stride length, step length, stride time, single limb support and stance (coordination function) by analyzing pressure between the foot plantar surface and the shoe insole. Therefore, this study developed an insole-type wearable pressure sensor.

In several other previous studies, the insole-type pressure sensor and smart shoes were developed for gait analysis and smart phone applications enabled real-time monitoring of the activities being carried out [27,28]. However, these devices are expensive, and the bio-signals collected and analyzed were reported to be not as accurate for use in clinical research [29]. To solve the cost-related problem of the sensors, researchers tried to apply pressure sensors for a gait analysis system, but they were found to be short-lived, making them still expensive for both users and researchers. Based on these advantages, researchers are conducting studies on an insole type textile pressure sensor considering the properties of conductive textile that is user friendly and inexpensive [30–34]. However, studies on the clinical application of the wearable textile insole sensor are scarce.

Therefore, this study developed a wearable textile capacitive pressure-sensing insole to test its feasibility in analyzing hemiparetic gait patterns and distinguishing its characteristics from normal gait.
