A Study on the Fabrication of Pressure Measurement Sensors and Intention Verification in a Personalized Socket of Intelligent Above-Knee Prostheses: A Guideline for Fabricating Flexible Sensors Using Velostat Film

Abstract

Intelligent transfemoral prostheses, which have recently been studied, are equipped with a microcontroller, providing appropriate motion functions for their walking environments. Thus, studies have been conducted to estimate user intentions in locomotion movements by applying biomechanical sensors inside the socket. Among them, a pressure sensor is used to determine the intentions of locomotion movements through changes in the internal pressure of the prosthetic socket. However, existing studies have a problem in that the reproducibility of pressure change data is degraded due to the non-detection and saturation of the pressure measurement value. Accordingly, this study proposes a fabrication method for a wide and flexible pressure sensor that can solve this problem and a method for the identification of user intentions in locomotion movements using it. The proposed system was fabricated with Velostat film, which has a smaller noise impact and can be fabricated in various sizes and shapes. The fabricated sensor was attached to four points inside the socket, confirming the possibility of detecting the intention of six movements according to the multi-critical detection method. The proposed pressure-sensor-based intention detection system can be applied individually by prosthetic users through simple tasks. Moreover, it will be universally applicable for commercialization.

1. Introduction

The transfemoral prosthesis is a device that assists in locomotion and daily life activities among femoral amputees [1,2]. Early transfemoral prostheses began in the form of no power source and later developed into a way of providing a power source to assist the user’s locomotion [3,4,5,6,7,8,9,10,11]. The transfemoral prostheses using a power source have been developed to be equipped with a microcontroller for determining various locomotion movements under a finite control state and provide appropriate motion functions for locomotion environments [12,13]. Accordingly, studies have been conducted together on methods for changing the way of the movement of prosthetic legs in response to changes in user intentions or locomotion environments [14,15,16,17]. Early studies used a method of predefined postures unrelated to the target movement to use changes in hip and knee joint angles as a trigger form or tapping the prosthetic leg in the form of a switch [18]. However, since this method causes unnaturalness unrelated to the target locomotion movement, a technique that makes a relatively natural change was required [19]. In response to this demand, studies have been conducted to monitor the user’s locomotion movement by detecting pressure changes inside the socket or EMG signals. Among them, the EMG signal acquisition method can implement a reliable algorithm within limited experimental conditions for a short time. However, the generalization of this method is not easy due to the problem of accumulating muscle fatigue [20,21]. The pressure change acquisition method is a principle of measuring the physical force acting on the outer wall of the socket according to the movement of the lower limb, and is relatively free from errors caused by the accumulation of muscle fatigue. Therefore, it is possible to compensate for the shortcomings of the EMG sensor for the purpose of detecting the user intention due to discontinuous pressure changes [22,23]. Based on these characteristics, studies have been conducted to detect pressure changes inside the prosthetic socket to determine the prosthetic user’s intention of changing the control method or monitoring locomotion movements [24]. A pressure sensor is attached to the inside of the socket and classifies the pressure change according to the movement in the form of a trigger or pattern [25]. Past pressure-sensor-based studies have mainly applied capacitive pressure sensors or resistive pressure sensors. A study using the capacitive pressure sensor fabricated a capacitive sensor in the form of a fiber or film [26]. Pressure data from the capacitive sensor according to the movement of the lower limbs could be classified into five movements, namely standing up, flat locomotion, stair climbing and descending, and ramp locomotion, based on the quadratic discriminant analysis method. However, the results of these studies reflect the method’s performance only in limited experimental environments, and capacitive sensors are vulnerable to skin impedance, which changes when the temperature rises inside the socket and sweating occurs. In addition, the capacitive pressure sensor has a slow recovery characteristic, with a full recovery time of approximately two seconds after pressure application. Since the average cycle of locomotion takes approximately two seconds, a pressure change occurs due to the movement of the lower limb before the full recovery of the sensor, affecting distinct pressure change data between each movement. As mentioned, the system using a capacitive pressure sensor shows appropriate pressure change performance within a limited experimental environment, but further studies are needed to apply it as a system suitable for the internal environment of the socket. In comparison, the resistive pressure sensor is relatively strong against the noise caused by a rise in temperature and sweat generation inside the socket, and the recovery time is relatively fast [27,28]. Therefore, it is considered that a resistance variable pressure sensor would be suitable for application inside the socket. Based on this, a related study has suggested the possibility of intuitive intention detection in the form of a trigger according to five movements by attaching an FSR (force-sensitive resistor) sensor, a film-type resistive pressure sensor, to the front and rear parts of the socket [29]. However, the FSR sensor applied in the above study was a small circular sensor with a diameter of approximately 13 mm. The small sensor could only check the pressure change in the local area when attached to the inside of the socket, making it difficult to verify the pressure change if a pressure application point occurs far from the attachment position. In the above study, the classification possibility of five lower limb movements was verified, but they were found to be influenced by the limited experimental environment in which the pressure application point and the sensor attachment position matched. The lower limb muscles in an amputated area have shorter muscle lengths and less muscle mass, as well as muscle contraction and relaxation, compared with non-amputated lower limbs. The reliability of pressure-change-based judgment data according to movement is reduced when the pressure change is checked only at a local part of the amputated lower limb since the reproducibility is reduced due to the lack of pressure change detection [21,22]. This problem causes the occurrence of a judgment error, which leads to a malfunction in prosthetic leg control, and there is a risk of a fall accident. This must be addressed in order to discern the user’s intention according to lower limb movement. In conclusion, it is necessary that the pressure sensor system applied inside the socket is not vulnerable to skin impedance and is made of a resistive variable material with fast recovery properties. In addition, it is necessary to apply it in a wide and flexible form to sensibly check the pressure change at the amputated lower limb. Therefore, this study proposes a pressure sensor system inside a prosthetic socket that can be fabricated in a wide single form and presents its fabrication guidelines. The proposed pressure sensor system is thin and flexible and aims to utilize a Velostat film, which can be fabricated as a pressure sensor through the attachment of electrodes. The application of a Velostat film has three advantages. First, the Velostat film is a resistance-change-based film that reduces electrical resistance when external pressure is applied. Thus, it is strong against the noise caused by body impedance changes compared with capacitive-based sensors. Second, it is possible to fabricate such a sensor in consideration of individual pressure ranges so that pressure non-detection and saturation problems do not occur because the measurement curve pattern and saturation range at the time of pressure application differ. Finally, since it is easy to fabricate in various shapes and wide sizes, there are no large fabrication costs due to a need for specific arrangements for measurement at an amputated area with low muscle mass. Based on these characteristics, the Velostat-film-based pressure sensor is expected to be able to solve the problems of the non-detection of pressure changes and saturation at a low cost and in a short time frame. However, studies on specific guidelines for the fabrication of pressure sensors using Velostat films are lacking. This study aims to present fabrication guidelines that can be customized according to the user based on Velostat film sensors. Sensors fabricated through the guidelines can be applied to the positions of the upper and lower rectus femoris and biceps femoris in order to detect pressure changes caused by lower limb movements in the forward and backward and upper and lower directions of the sagittal plane. A sensor fabricated through this process will address the limitations of the existing socket internal pressure sensor system. In other words, it is possible to apply a sensor that is simple to fabricate and has excellent detection performance due to less noise against the skin reaction when applied inside the socket. In addition, by cutting the outside of the sensor in a V shape, it is possible to obtain a result that can be applied inside a wide and curved socket rather than a general small plane. In Section 3, the possibility of obtaining an improved sensor, as proposed in this study, is verified through comparative experiments with the existing capacitive pressure sensor. The proposed pressure-sensor-based intention detection system can be applied individually by prosthetic users through simple tasks, and universal applications for commercialization will be possible.

2. Materials and Methods

In this section, the method of fabricating a flexible and wide pressure sensor using a Velostat film and the fabrication guidelines are presented. In addition, the pressure measurement principle and walking intention classification technique of the fabricated pressure sensor are presented. Section 2.1 presents the sensing principle, fabrication method, and structure of the Velostat-film-based pressure sensor, which can be customized according to the pressure range inside the socket. In Section 2.2, the method of selecting the size of the sensor to apply it inside the socket and the process of cutting it to be applied to the bending socket are presented.

2.1. Velostat-Film-Based Pressure Sensor with Individual Application of Lower Limb Pressure Ranges

The Velostat film is a material that changes the electrical resistance and current by compressing the conductive layer when external pressure is applied. Therefore, the measurement of the changing resistance can confirm the magnitude of the applied pressure [30,31,32]. Figure 1b is a graph showing the pattern of resistance change when pressure is applied to a Velostat film of three different sizes. The larger the size of the Velostat film, the smaller the initial resistance value, which reflects the different characteristics of the resistance change at the time of pressure application and the saturation point reaching the minimum resistance. Therefore, if sensors of various sizes are fabricated using the Velostat film, different pressure change characteristics can be considered depending on the sensor attachment point. Figure 1c shows the method of fabricating a pressure sensor with a Velostat film. The sensor is fabricated by placing an electrode foil on both sides of the Velostat film and coating the outer surface with an adhesive film. This shape is similar to a sandwich structure. Figure 1d represents a voltage distribution circuit and a Wheatstone bridge circuit for the determination of the magnitude of the applied pressure.

Since the resistance value of the Velostat film changes according to the applied pressure scale, the scale of the applied pressure can be obtained by constructing a circuit using the fabricated sensor as a factor. As the sensor is fabricated using a voltage distribution circuit, the amount of applied pressure can be obtained through the output voltage value according to Equation (1). The voltage distribution circuit may be advantageously used in a relatively simple configuration. The Velostat sensor consists of an upper part close to the applied voltage, and the internal reference resistance consists of a lower part close to the ground in order to verify the upward output in the pressure application. The internal reference resistance was configured to present the output of 500 mV in a state in which pressure was not applied. The possibility of applying the Velostat-based sensor inside the socket was confirmed through the previous research conducted by our research team [33].

$$V_{out}=V_{cc}\times \frac{R_{reference}}{R_{velostat}+R_{reference}}$$

(1)

2.2. Individual Application Method according to User Attachment of the Fabricated Sensor

This section describes a customized application method according to the user of the Velostat-film-based pressure sensor.

Table 1. Fabrication and application method of the Velostat-film-based customized sensor.

1. Check the individual pressure range inside the socket and select the fabrication size of the sensor

2. Cut the outer part of the sensor in accordance with the flexion of the attachment area

3. Attach the sensor and wear the prosthetic leg

4. Detect the user intention according to the movement of the lower limb

shows the process of its application. First, it is necessary to check the individual pressure range inside the socket and select the fabrication size of the sensor. When the size is selected, it is necessary to cut the outer part to prevent wrinkles caused by the flexion of the attachment area. The cut sensor is attached to a position where the movement of the lower limb can be checked. The user intention is detected according to the lower limb movement in the attached sensor. The process of these four steps is described in detail below.

2.2.1. Check the Individual Pressure Range inside the Socket and Select the Fabrication Size of the Sensor

The range of pressure changes occurring inside the socket varies depending on the user and the sensor attachment position. Sensors that do not consider the pressure range carry problems such as undetected pressure changes and the saturation of pressure measurements. Therefore, when applying a pressure sensor inside the socket, the individual pressure range at the application position must be verified. The Velostat film has different resistance change characteristics depending on the size. The process of selecting an appropriate sensor size by checking the pressure range is as follows. Figure 2a shows the pressure range inside the socket. The value when the maximum load is applied in the direction of the ground after wearing the prosthetic leg is $P_{max}$, and the value when the load is removed by raising the lower limb is $P_{min}$. It is possible to consider the pressure range by selecting the size showing linear pressure change characteristics within the range above $P_{min}$ and below $P_{max}$. Since the pressure value is measured by A/D conversion, it is based on the reference voltage of 3.3 V.

For pressure discrimination, a range of 2.8 V to 3.3 V is set as the maximum pressure and a range of 0.5 V to 1 V is set as the minimum range. The pressure change due to the movement of the lower limb is verified to be a change that can be distinguished only when a voltage change of 1 V or more occurs. The size and shape of the sensor for application are presented as a production guideline through pressure measurement experiments in Section 3.

In a standing state, there is an initial pressure value inside the socket due to the load of the body [34]. Therefore, whether the pressure changes due to the movement of the lower limb should be determined based on the initial pressure. Figure 2b shows how to convert the change in pressure based on the initial pressure. It is necessary to set the initial pressure $P_{init}$ presented by the body load in the standing state to the reference value of 0% and convert the maximum pressure $P_{max}$ to 100%, and the minimum pressure $P_{min}$ to a percentage of −100%. The conversion value of the pressure change during the lower limb movement follows Equation (2). If the measurement pressure is greater than the initial pressure, a positive value is shown through the upper expression of Equation (2) because of the positive change. When the measurement pressure is less than the initial pressure, a negative value appears through the lower equation of Equation (2). The proposed conversion method is applied to the individual pressure range of each sensor, thereby completing the initial setting of the sensor.

$$\begin{array}{c}p=\frac{p_{mea}-p_{init}}{p_{max}-p_{init}}\times 100 (p_{mea}>p_{init})\\ p=-100+\frac{p_{mea}-p_{init}}{p_{min}-p_{init}}\times 100 (p_{mea}<p_{init})\end{array}$$

(2)

2.2.2. Cutting the Outer Part of the Sensor according to the Flexion of Attachment Areas

Professor Noor A. Osman et al. aimed to verify pressure changes according to five locomotion movements by attaching a small pressure sensor inside the socket [28]. The small-sized sensor applied by the research team was located only in a local area of the amputated muscle, with reduced muscle contraction and relaxation ability. As a result, reproducibility was degraded when the position of the lower limb changed inside the socket. In conclusion, sensors that detect intention based on the movement of the lower limb inside the socket should be applied in a large area rather than a narrow and specific region. However, a large-sized sensor causes an unfastened state in which it does not adhere to the inner surface of the socket, where flexion exists. The unfastened state reduces the reproducibility by causing shape deformation, such as wrinkles on the sensor. This problem can be solved by cutting the outside of the sensor. The wider the cut shape, the easier it is to resolve the unfastened state.

However, this causes the excessive loss of the sensor area, as shown in Figure 3b. The optimal amount of cutting, with little loss of the sensor and without the occurrence of the unfastened state, as proposed in this paper, is within 10% of the total sensor area. If cutting is carried out in excess of 10%, there is a risk of not detecting pressure changes due to the loss of the sensor area. Figure 3c shows a triangular cut shape to resolve the unfastened state. Figure 3d shows the results of a simple experiment comparing the pressure measurement pattern according to the presence or absence of a sensor cut. If an uncut sensor is attached to a curved surface, the conductive layer of the sensor is compressed, and an overpower value exceeding the initial power range appears even when the pressure is not applied. This reduces the detection resolution of the sensor.

In comparison, the cut sensor has a low initial measurement voltage of 1 V or less, resulting in a wide range of pressure changes. Through this, it can be confirmed that the cutting of the sensor is necessary according to the flexion of the attachment area. The minimum area of cutting according to the flexion section is presented as a production guideline through experiments in Section 3. Figure 3e shows the results of the experiment to determine whether the pressure measurement characteristics are maintained according to the presence or absence of cutting in the outer part. The pressure measurement characteristics of the Velostat film pressure sensor, which was processed by cutting the outer part, were maintained at the same values as before the cutting. To verify this, cutting was performed on only one of the two sensors of the same size, and an experiment was conducted to compare the pressure measurement characteristics by applying the same pressure. Figure 3e shows the results of comparing the pressure measurement characteristics of the uncut and cut sensors fabricated at the same 2500 mm² size. The cut was composed of a triangle, 5 mm wide and 10 mm long, on each of the four sides of the sensor.

The measurement of these data was performed by sequentially applying pressure within the average pressure range of 200 kPa inside the socket. It was confirmed that the pressure measurement characteristics of the Velostat-based pressure sensor were maintained even when cutting the outer part. This sensor can be applied inside the socket to resolve the unfastened state with the same pressure characteristics.

2.2.3. Attaching Sensors and Wearing the Prosthetic Leg

The change in pressure inside the socket is caused by a change in the position of the load application point, where the weight of the prosthetic leg distributed according to the movement of the lower limb is concentrated to a specific point, and the physical force application point acting on the outer wall of the socket [20].

Figure 4a shows the position of the sensor to be attached inside the socket. The sensor is intended to be applied to the upper and lower rectus and biceps femoris, which are the positions presented in previous studies [35]. The four positions are suitable for the detection of pressure changes caused by movements of the lower limb in the forward and backward and upper and lower directions of the sagittal plane. The pressing force from the standing position of the prosthetic user due to gravity is shown in Figure 4b, and the pressing force from the position where the load is removed by lifting the lower limb at a low level is shown in Figure 4c. Based on this basic principle, it is possible to detect the user’s intentional lower limb movement in the event of discontinuous pressure changes.

2.2.4. Detection of the User’s Intention according to Movements of the Lower Limb

Section 2.2 describes the sensor selection process according to the individual pressure range inside the socket and the process of cutting the outer part of the sensor according to the flexion of the attachment area. After attaching the sensor fabricated through this process to four target locations inside the socket, the goal is to detect the intention to change the control method according to discontinuous movements of the lower limb. The lower limb movements to be detected in this study are six basic movements that occur frequently in daily life: standing up, locomotion, standing to sitting, sitting to standing, stair climbing, and stair descending.

Since the six movements have no similarity in their locomotion sequences, it is possible to classify them according to their changing patterns based on the initial pressure presented in the standing position [26]. The six suggested movements are classified as pressure changes occurring at the start of the transition to each posture. Among them, in the case of stair locomotion, the user requires power before performing the movement. Thus, the knee joint bending movement, similar to stair climbing, and the weight application posture, similar to stair descending, are defined and detected in advance. Figure 5 shows the prediction of the pressure change pattern in the socket that will occur during the movement of each lower limb.

Figure 6 represents a flowchart illustrating the process of detecting pressure changes according to each movement and classifying the pressure changes into intention signals. The proposed pressure sensor system receives pressure changes from four locations inside the socket during the movement of the lower limb. The input changes are converted into a ratio form based on the initial pressure. The converted pressure change rates at the four locations of the sensor are classified into six movement intentions through a multi-critical scheme. The suggested movement intention determination method will enable the simple and intuitive detection of the movement intention.

3. Experiments and Results

In this section, an experiment is described that verifies the possibility of detecting the user intention using the proposed sensor system. Section 3.1 examines the pressure measurement characteristics according to the fabrication size of the Velostat film sensor and presents the fabrication guidelines for the sensors, which can be applied individually according to the user and the attachment points. In Section 3.2, the fabricated sensors are applied inside the socket to check the possibility of detecting user intentions according to the movement of the lower limb.

3.1. Guideline for Fabricating Pressure Sensors for Individual Applications

This study examines the pressure measurement characteristics of the Velostat film sensor according to the fabrication size and presents the fabrication guidelines for the sensors, which can be applied individually depending on the user and the attachment points.

Figure 7a shows eight sensors fabricated by increasing the length by 10 mm to check the pressure change pattern according to the size of the Velostat film. Figure 7b shows the experimental state when pressure is applied to the sensor using a force gauge. The fabrication sensor was sequentially pressurized below the average pressure range of 200 kPa inside the socket [36,37,38]. Figure 8 shows the range of pressure changes shown by the pressure application and its aspect. A suitable pressure range for the internal application of the sensor is one in which it does not cause saturation and exhibits linear properties. Therefore, in this stage, the range with linear characteristics was identified before saturation occurred, and this is shown in

Table 2. Size of the Velostat-film-based sensor is suitable for application depending on the pressure range inside the socket.

Range [kPa]	Sensor Size [mm2]	Considerations for Application
Under 50	400, 600	Narrow area and pressure range allow for large changes in small pressure. Suitable for use in areas with small muscle mass or for children and low-weight users.
Under 100	800	Suitable for applications to areas with pressure generation of 100 kPa or less to users.
Under 150	900, 1000, 1200, 1600	Suitable for applications to areas with pressure generation of 150 kPa or less to users.
Under 180	2500	Large area and pressure range: Suitable for use in areas that require a detailed check of muscle movement or in users with high weight and high activity.

. With the fabrication of the sensor with reference to the pressure range selection guidelines presented in Table 2, it will be possible to apply the sensor considering individual pressure ranges. The shape of the fabrication sensor proposed in this study is rectangular, but various types of sensors, such as circular, elliptical, and diamond-shaped sensors, can also be fabricated depending on the internal structure of the socket.

This section aims to identify the unfastened state that occurs when 2500 ${\mathrm{m}\mathrm{m}}^2$-sized sensors are attached to four flexion surfaces in order to present an individual sensor cutting method according to the socket flexion. Figure 9 shows the experimental results of determining the unfastened state according to the flexion radius and the amount of cutting of the outer part. Data at the top of the graph are the initial output values identified by attaching sensors to four flexion surfaces with different flexion radii. The unfastened state of each sensor can be confirmed through the initial output value of the sensor. The initial output value when the pressure is not applied is 500 mV, set through the internal reference resistance. Therefore, if a measurement value exceeding 500 mV is verified, the bending or crumpling of the sensor occurs due to the unfastened state.

The outside of the sensor was cut into four sizes, and each was attached to the flexion surface to identify the minimum cut amount in which the initial output value close to 500 mV was measured. Based on data at the bottom part of Figure 9, the decrease in the unfastened state when applying the sensor cut into the minimum area can be confirmed, and this is shown in

Table 3. Minimum cut according to the radius of flexion.

	R0	R60	R80	R100	R120
Pictures
Image
Cutting amount [mm]	Base 0 (0%) Height 0 (0%)	Base 7 (14%) Height 14 (28%)	Base 6 (12%) Height 12 (24%)	Base 5 (10%) Height 10 (20%)	Base 4 (8%) Height 8 (16%)
Cutting rate [%] (Loss Rate)	0	7.84	5.76	4	2.56

. If cutting is performed by referring to the cutting guideline according to the flexion radius presented in Table 3, it will be possible to apply the sensor considering the flexion inside the socket of an individual user. If sensor cutting other than the size suggested is required, individual cutting according to the flexion radius is possible by applying the cutting ratio suggested in this study.

3.2. Experiment and Performance Verification according to the Application Process of the Lower Limb Movement Intention Detection System

In this section, the sensors are fabricated and applied according to the fabrication guidelines of the proposed lower limb movement intention detection sensor, and this verifies the intention detection performance of the fabricated system.

3.2.1. Configuration of the Velostat-Film-Based Pressure Sensor System

Data from the pressure sensor inside the socket proposed in this study are to be obtained by fabricating an analog data collection module of 4CH. Figure 10 shows the configuration and specifications of the fabricated system. The fabrication module consists of two small layers, 45 mm wide and 43 mm long, to prevent interferences in movement due to attachment to the prosthetic leg.

The lower surface consists of a Bluetooth FB301 device for wireless data transmission, receiving power through a 3.7 V battery, and the upper surface consists of ADC for the acquisition of 4CH pressure sensor data, an SD card slot to store acquired data, and two USARTs for data transmission and reception. For high-speed communication, the communication speed of USART is set at 57,600 bps, and pressure data are acquired every 10 ms. Acquired data are transmitted to a PC system and recorded in its storage device at the same time. The data acquisition module uses a 32-bit microcontroller that supports 12-bit ADC for high-resolution analog data acquisition and user intention determination. Pressure sensor data of 4CH will be converted through ADC, and acquired sensor data will be classified according to the user intention through MCU.

3.2.2. Experimental Participant Information

In order to conduct the experiment according to the application process of the lower limb movement intention detection system, one male amputee, one male non-amputee, and one female non-amputee with various physical conditions were selected. Subject 1 was a rectus amputee, and Figure 11a shows the appearance of Subject 1 when wearing the prosthetic leg. The socket worn by Subject 1 was a boa-type socket designed to increase adhesion by making four cuts to the front, rear, left, and right sides to fix the socket and the amputated lower limb. Figure 11b shows the simple experimental prosthetic leg and a diagram showing the sensor attachment position of the experimental prosthetic leg. The sockets worn by Subjects 2 and 3 were simple experimental prosthetic legs fabricated for experiments based on the wearing of prosthetic legs by non-amputated subjects. The experimental prosthetic leg had a structure that was not able to identify pressure changes at the lower part of the biceps femoris. Therefore, the sensor at the lower part of the biceps femoris was attached to the lower part of the knee joint to check the change in pressure in the vertical direction occurring at the lower part of the biceps femoris. Figure 11c shows the prior information on the three subjects.

3.2.3. Experiments according to the Application Process of the Lower Limb Movement Intention Detection System: Initial Sensor Application Step

Since the pressure range inside the socket varies depending on the physical characteristics of each subject, it is necessary to apply appropriate individual sensors. To this end, standing up, maximum pressure application, and minimum pressure application were applied, as shown in Figure 12, in order to determine the pressure pattern inside the socket of each subject. The size of the pressure sensor suitable for the individual pressure range was selected, and the cutting ratio was designated according to flexion to prevent an unfastened state from occurring inside the socket. The sensors fabricated at the specified size and cutting ratio were attached to the inside of the socket to configure an individual in-socket sensor system.

Table 4. Configuration of the individual in-socket sensor system in subjects.

	Subject 1	Subject 2	Subject 3
Physical Conditions	Height 171 cm, Weight 82 kg, Male Rectus Amputee	Height 155 cm, Weight 50 kg, Female Non-Amputated	Height 174 cm, Weight 75 kg, Male Non-Amputated
Selected Sensor [mm2]	Sensor 1 1600 (about 150 kPa)	Sensor 1 800 (about 100 kPa)	Sensor 1 1600 (about 150 kPa)
	Sensor 2 1600 (about 150 kPa)	Sensor 2 800 (about 100 kPa)	Sensor 2 800 (about 100 kPa)
	Sensor 3 2400 (over 180 kPa)	Sensor 3 2500 (over 180 kPa)	Sensor 3 2500 (over 180 kPa)
	Sensor 4 2400 (over 180 kPa)	Sensor 4 2500 (over 180 kPa)	Sensor 4 2500 (over 180 kPa)
Cutting	Sensor 1 Base 8%, Height 16%	Sensor 1 Base 8%, Height 16%
	(flexion radius about 120)	(flexion radius about 120)
	Sensor 2 Base 10%, Height 20%	Sensor 2 Base 10%, Height 20%
	(flexion radius about 100)	(flexion radius about 100)
	Sensor 3 Base 10%, Height 20%	Sensor 3 Base 14%, Height 28%
	(flexion radius under 60)	(flexion radius under 60)
	Sensor 4 Base 8%, Height 16%	Sensor 4 Base 8%, Height 16%
	(flexion radius about 120)	(flexion radius about 120)
Image

summarizes the sensors selected according to the physical characteristics of the subjects and the pressure range and flexion of each attachment part.

Subject 2 showed a pressure range of around 100 kPa when the maximum pressure was applied due to their relatively small weight, and four sensor sizes were selected accordingly. Subject 3 showed a pressure range of 150 and 100 kPa at the top and bottom of the front, respectively, and the sensor size suitable for the maximum pressure of 180 kPa or more was selected at the lower and rear parts. Subjects 2 and 3 showed the same flexion radius at the attachment point because they used the same simple experimental prosthetic leg for non-amputated subjects. Compared with the lower part of the socket, the upper and rear parts of the front were cut at 8% of the lower part and 16% of the height to suit the gentle flexion radius of around 120. The cutting was performed by cutting the lower part at 14% and the height at 28% to prevent an unfastened state in the lower part, with a small bending radius of 60 or less. In the case of the rear lower limb of Subject 1, sufficient muscle mass did not occur compared with the initial pressure caused by the initial physical load. For this reason, in order to check the pressure change in a broadened area, an additional 60-mm-wide and 40-mm-long sensor with a different shape compared with the sensors selected for Subjects 2 and 3 was fabricated and applied to the upper and lower positions of the rear. It was designed to eliminate the unfastened state by performing cutting in a total of six locations, two in the horizontal direction and one in the vertical direction, to suit the broadened range.

In this study, a Velostat-film-based pressure sensor system is proposed to solve the problems of the non-detection of pressure changes and pressure saturation identified in previous studies. In order to verify whether the proposed system solves these problems, the sensor of Subject 2 and the capacitive sensor applied in the method proposed in this study were applied inside the socket in order to compare the pressure change patterns according to locomotion movements. Figure 13 shows the results of comparing the pressure changes in the sensor according to the locomotion of two steps in Subject 2. Section (1) of Figure 13 represents the swing phase within the locomotion cycle, and section (2) represents the stance phase. Figure 13a shows the capacitive sensor presented in the locomotion, and Figure 13b presents the sensor proposed in this study. Capacitive sensors are vulnerable to the skin impedance, which changes when the temperature rises inside the socket and sweating occurs; they have slow recovery characteristics and are applied only to local areas with a size of 4 ${\mathrm{c}\mathrm{m}}^2$. In addition, the prosthetic user and the individual pressure range inside the socket were not considered. As a result, as shown in Figure 13a, it was verified that a discontinuous step occurred due to the slow recovery in the lower sensor of the rectus femoris in locomotion movement. In addition, at the lower part of the biceps femoris, since a small-sized sensor was applied to the area with less muscle mass after the saturation of the measured value, the load application points did not match, and thus no pressure change was confirmed.

In comparison, the Velostat film sensor, considering the pressure range proposed in this study, did not show a discontinuous step in locomotion movement. It was confirmed that the measurement problem identified in Figure 13a was solved by fabricating a resistance-change-type sensor, rather than a capacitive type, to cover a wide area and by ensuring the consideration of the individual pressure range.

3.2.4. Experiments according to the Application Process of the Lower Limb Movement Intention Detection System: Performing Six Locomotion Movements

An experiment was conducted using the process presented in

Table 5. Experimental process.

(1) Standing

(2) Level locomotion in a standing posture

(3) Performing knee joint flexion in a standing posture (predefined posture for moving upstairs)

(4) Load application in a standing posture (predefined posture for moving downstairs)

(5) Sitting in a standing posture

(6) Standing in a sitting posture

in order to confirm the possibility of detecting the movement intention of the proposed pressure sensor system. This experiment was conducted in the following order. A standing position is presented in (1). It is intended to confirm the conversion time to start flat locomotion in the initial standing state in (2). Stairs are to be first recognized in the stair locomotion movements in (3) and (4), and then power is applied; pressure changes are confirmed by defining knee joint bending movements similar to stair climbing and weight application, with a posture similar to stair descending. The transition from a standing state to a sitting position is carried out, as shown in (5). The transition from a sitting position to a standing state is presented in (6).

Figure 14 shows the results of pressure changes according to the proposed six locomotion movements. When the movement is switched from a standing state to a flat locomotion state, the load is removed, and the pressure drops below −50% from the upper part of the rectus femoris. At the lower part of the rectus femoris, the weight of the prosthetic leg is concentrated. Then, it is possible to verify that the pressure change increases by more than 50% at the upper part due to the pressing of the inclined lower limb. During stair climbing, the result of performing the predefined knee joint bending movement appears. The upper part of the biceps femoris removes the pressing force acting on the outer wall of the socket, causing the pressure to drop to −100%. At the lower part of the biceps femoris, the pressing force increases, and the pressure rises to +50%. The lower part of the biceps femoris shows a decreasing pressure change in common but is divided into a decreasing result after rising and an immediate decreasing result. This difference is determined to be due to the individual user’s habit of performing the knee-bending posture. In this case, it is not designated as a threshold value because it is difficult to identify a clear pressure change. During the stair descending movement, the result of performing the predefined intentional weight application movement appears. The weight application movement increases the pressing force, showing an elevated form in all four positions.

It can be seen that the pressure rises to more than 50% in the upper part of the rectus femoris and 100% in the lower part of the rectus femoris and the biceps femoris, entering a saturated state. In the sitting movement, the lower limb is inclined in the front and lower directions, accompanied by knee joint bending. It can be seen that the pressure decreases to −100% at the upper part of the rectus femoris, with an increase to +50% at the lower part of the rectus femoris caused by a large pressing force and an increase to +50% at the lower part of the biceps femoris due to a small pressing force. In the conversion movement from the sitting posture to the standing posture, the pressure change based on the sitting posture should be checked. Due to the low load on the lower part of the seated position, the sensors in the upper position of the rectus femoris and the upper and lower positions of the biceps femoris represent a low-pressure condition of −50%. In the conversion movement from a standing to a sitting position, the pressure change increases to 0% or more at the upper part of the front and rear parts and +50% or more at the lower part of the rear parts. In the case of the sensors in the position of the biceps femoris, the pressure value generated due to the pressing of the lower part in the seated state is different. Therefore, two positions of the biceps femoris exceed the threshold setting.

Table 6. Pressure change patterns according to the movement of the lower limb intention.

	Pressure Change of Sensor 4 [%] (Upper of Biceps Femoris)	Pressure Change of Sensor 3 [%] (Lower of Biceps Femoris)	Pressure Change of Sensor 2 [%] (Lower of Rectus Femoris)	Pressure Change of Sensor 1 [%] (Upper of Rectus Femoris)
Standing	0	0	0	0
Walking	Rising 0→50	-	Rising 0→50	Decreasing 0→$-$50
Up Stair	0	-	Rising 0→50	Decreasing 0→$-10$0
Down Stair	Rising 0→100	Rising 0→100	Rising 0→100	Rising 0→50
Stand to Sit	0	Rising 0→50	Rising 0→50	Decreasing 0→$-10$0
Sit to Stand	Rising $-5$0 or less→0	Rising $-5$0 or less→0	-	Rising $-5$0 or less→0

summarizes the internal pressure change patterns of the socket to set a threshold based on the results of the pressure changes according to the six movements presented in Figure 14. In the standing position, all four sensors begin their initial state at 0%, and when converted to flat locomotion, they are −100%, 100%, −100%, and 50% in order from the front to the rear. When converting to the stair climbing movement, the results are −100%, 100%, not set, and 0%, and when converting to the stair descending movement, the value shows a 100% rise at all four positions. Conversion from standing to sitting presents values of −100%, 100%, 50%, and 0%, and for conversion from sitting to standing, the value changes from −50% to 0%, not set, 50%, and 0%. It is possible to detect the movement of the lower limb to be switched by setting a threshold value based on the suggested pressure change characteristics. It was confirmed that the pressure change patterns according to each lower limb movement measured through the application of the fabricated sensor were similar to the expected pressure changes according to the six lower limb movements presented in Section 3. In the experimental results, there is a position at which different pressure changes are determined depending on the subject, even though the same behavior is verified. It is difficult to set a clear threshold at the lower part of the biceps femoris in flat locomotion and stair climbing movements and the lower part of the rectus femoris in the conversion from sitting to standing movements. This is caused by a difference in the structure of the transfemoral prosthesis socket for each subject. In addition, it is estimated that the difference occurs due to the variations in the subjects’ physical structure, posture, and habits of performing movements. For this reason, when setting the scale of the pressure change at each sensor position to the critical value, only the common occurrence conditions are verified and defined.

Through these experiments, the possibility of measuring pressure changes according to the lower limb movement of the flexible pressure sensor system based on the Velostat film proposed in this study was confirmed. It was verified that accurate classification was possible according to each movement by addressing the problems of not detecting pressure changes and the saturation of existing pressure sensors. In addition, the possibility of classifying the movement intention was determined by confirming the patterns of pressure changes according to six intentional movements. There was a point at which the time required for a pressure change for each subject was different, but this could be addressed according to the degree of movement learning of the user. The Velostat-film-based sensor proposed in this study is able to set a relatively clear threshold by solving the problems of not detecting pressure changes and the saturation of existing capacitive sensors.

4. Discussion

This study proposed a Velostat-film-based pressure sensor system to solve the problems of the non-detection of pressure changes and saturation of measured values in the existing pressure change measurement system to determine the locomotion movement intentions of transfemoral prosthetic users. The intention detection system based on the measurement of internal pressure changes in the socket, which was previously studied to detect the intentions of locomotion movements of prosthetic users, presents a problem in terms of not detecting pressure changes in an amputated section with low muscle mass. In addition, there is a problem in which the linear sensing pressure range of the sensor is lower than the internal pressure range of the socket, resulting in the early saturation of the pressure measurement. In this study, it was possible to fabricate a wide type of sensor. A pressure sensor system designed to eliminate the problems of not detecting pressure changes and the saturation of measured values is proposed, and its fabrication guidelines are presented. The system proposed in this study has a different pressure measurement range depending on the size; it is based on the Velostat film, which can be fabricated in various sizes and shapes and is fabricated in consideration of individual pressure ranges depending on the user. Through this study, it was confirmed that a sensible pressure change of 1 V or more occurred due to the elimination of the problem whereby the pressure change cannot be measured in an area with low muscle mass. In addition, it was verified that a linear output was generated within the required measurement range by addressing the problem of the early saturation of pressure measurements due to the low linear sensing pressure range of the sensor. A pressure sensor with the improved detection of pressure changes and the reduced early saturation of measured values was applied to four positions inside the socket to confirm the possibility of detecting six movements: standing up, locomotion, stair climbing, stair descending, sitting, and standing. However, since its durability has not been verified, it is necessary to confirm its long-term availability in the future. In addition, the threshold value of the operating signal defined in this study will need to be converted into a standardized equation to increase the classification performance.

5. Conclusions

In this study, the fabrication of a wide and flexible pressure sensor capable of detecting pressure changes inside the socket and solving the problem of the saturation of measured values and a technique to identify the intention of the user were proposed. The proposed system is composed of Velostat film, which is less affected by the noise caused by changes in body impedance and can be produced in various sizes and shapes. The sensor fabricated through the proposed method addressed the problems of the non-detection of pressure change and early saturation. In addition, it was attached to four locations inside the socket to confirm the possibility of detecting the intention of six actions according to the multi-threshold detection method. The proposed pressure-change-based user movement intention detection system can be expanded into a system that considers the possibility of individual fabrication and universalization in combination according to the user’s lower limb. In addition, it does not require the additional fabrication of a dedicated socket in a form that can be attached to the existing user’s socket. Thus, it is expected to offer excellent benefits in terms of its fabrication cost.