This article describes a technique for applying vibrotactile stimuli to the thigh of a human participant, and measuring the accuracy and reaction time of the participant's volitional response for various combinations of stimulation location and frequency.
Artificial sensory feedback (ASF) systems can be used to compensate for lost proprioception in individuals with lower-limb impairments. Effective design of these ASF systems requires an in-depth understanding of how the parameters of specific feedback mechanism affect user perception and reaction to stimuli. This article presents a method for applying vibrotactile stimuli to human participants and measuring their response. Rotating mass vibratory motors are placed at pre-defined locations on the participant’s thigh, and controlled through custom hardware and software. The speed and accuracy of participants’ volitional responses to vibrotactile stimuli are measured for researcher-specified combinations of motor placement and vibration frequency. While the protocol described here uses push-buttons to collect a simple binary response to the vibrotactile stimuli, the technique can be extended to other response mechanisms using inertial measurement units or pressure sensors to measure joint angle and weight bearing ratios, respectively. Similarly, the application of vibrotactile stimuli can be explored for body segments other than the thigh.
Artificial sensory feedback (ASF) can be defined as the practice of providing real-time biological information to individuals, often compensating for compromised proprioception or other sensory mechanism. ASF has been long used in the realm of rehabilitation of injured or disabled persons to assist in recovering of aspects of physical function and movement1–3, allowing individuals to control physical processes that once were an involuntary response of the autonomic nervous system4. A subcategory of ASF, biomechanical biofeedback, uses external sensors to measure parameters relating to balance or gait kinematics, and communicate this information to the individual through some sort of applied stimulus. An increasingly popular approach to biomechanical feedback employs small vibrating motors, or tactors, placed at different parts of the body to provide spatial as well as temporal feedback. Previous literature has showed promising results supporting the use of vibrotactile feedback in applications to individuals with lower-limb amputations, vestibular impairments, and aging-related loss of balance5–9.
A thorough understanding of the mechanisms controlling an individual's perception and response to specific stimuli is necessary for informing effective implementation of ASF systems for different applications. For vibrotactile feedback, chief among these mechanisms are proprioception and the sensorimotor response, specifically the user sensitivity to the applied vibrations and the time required to execute the desired reaction. Any sensory information communicated through vibration stimuli must be encoded as specific combinations of vibration frequency, amplitude, location, and sequence. Therefore, design of vibrotactile ASF systems should select combinations of parameters to maximize user's perception and interpretation of the stimuli, as well as the timeliness and accuracy of the resulting motor response. The goal of this protocol is to provide a platform from which to evaluate response times and response accuracy to various vibrational stimuli to inform the design of ASF systems for use with different sensory-impaired populations.
The methods described here builds on prior research exploring human perception of tactile and vibrotactile feedback3,5,6, and was developed for use in two previous studies10,11. The latter two studies employed this protocol to examine the effects of vibration frequency and location on the accuracy and timeliness of user responses in lower-limb amputees, showing that both parameters significantly affect the outcome measures, and that a high degree of response accuracy can be achieved. These results can be used to inform the ideal placement of tactors in future studies and clinical applications of vibrotactile ASF systems. Other recent work by Crea et al.12 examined user sensitivity to changes in vibration patterns applied to the thigh during walking, using verbal responses to signify perceived changes to the vibration patterns, rather than a motor response. While these verbal responses can be used to measure detection accuracy, they do not account for errors and delays that may be present in the motor control process.
The primary setup for the following experiments consists of a number of vibrating motors connected to pulse-width-modulated output pins of a microcontroller board. The board is, in turn, controlled through a Universal Serial Bus (USB) connection to a computer running commercially available system design software. The motors require an additional amplifying circuit to ensure sufficient voltage and current is supplied over a wide range of vibration frequencies. An example amplifier circuit is shown in Figure 1. The bipolar junction transistor (BJR) in the figure can be replaced with smaller metal-oxide-semiconductor field-effect transistor (MOSFET) for more efficient operation and smaller size. Similarly, the entire amplifying circuit can be replaced by an off-the-shelf haptic motor driver to provide additional control and reduced size. Each motor requires its own circuit, and using the equipment listed in this paper, up to ten motors can be controlled by a single microcontroller board.
Figure 1. Motor Wiring. (A) The amplification circuit for a single vibration motor is shown. Each motor requires a separate circuit and must be connected to a unique PWM output port on the microcontroller. The VDD here represents the 3.3 V power supplied by the microcontroller board, and the resistor R2 serves as a pull-down resister to ensure the transistor switch remains open when zero voltage is applied. (B) An example of the physical wiring of two motors. Although eight individual amplification circuits are shown, only two are connected to vibration motors. In this protocol R1 = 4.7 kΩ and R2 = 100 kΩ. Please click here to view a larger version of this figure.
The following protocol was approved by the Research Ethics Board at Holland Bloorview Kids Rehabilitation Hospital.
1. Motor Calibration
Figure 2. Accelerometer Mounted to Motor. The tri-axis accelerometer (green) is mounted to the coin motor with its z-axis orthogonal to the flat surface of the motor for calibration. Each motor was activated using different duty cycles, and the corresponding vibration frequencies were recorded by the accelerometer. Please click here to view a larger version of this figure.
2. Placing the Motors
Figure 3. Test Platform for Experiments. A custom test platform was built to house the microcontroller boards and push buttons. Motors can be attached directly to the skin (as shown), or with a prosthetic liner between the motor and skin. Please click here to view a larger version of this figure.
3. Experiment 1: Applying Stimuli and Recording Reaction Time
4. Experiment 2: Distinguishing between Stimuli
Note: This experiment can be conducted entirely independently from Experiment 1. A single motor or multiple motors can be used. The specific locations of the motors can vary depending on the application and research questions.
Figure 4 illustrates the calibration curves identifying the PWM value for a 180 Hz vibration frequency of a single motor. Starting at a 50% duty cycle, the PWM values are iterated until the primary frequency spike occurs at 180 Hz. Successful calibration trials should show a clear spike at the primary vibration frequency. Poor fixation of the accelerometer to the motor, or of the motor to a support surface may result in a more diffuse FFT without a clear spike. In this situation, the calibration trial should be repeated after the mounts have been adjusted to ensure a better connection.
Figure 5A shows reaction times between stimulus and push button response recorded for Experiment 1 for three vibration frequencies, 140 Hz, 180 Hz, and 220 Hz, applied to the anterior surface of the thigh for ten able-bodied participants and three amputees10. Repeated measures analysis of variance (ANOVA) and Tukey post-hoc analysis using the Bonferroni correction was used to identify the specific effects of each frequency. These data show a relatively tight spread of data for each frequency in the able-bodied population, and a significant frequency effect. Reaction times for distinguishing between pairs of vibration frequencies are shown in Figure 5B, and can be analyzed using the same procedure as the single frequency tests. Similar analyses can be conducted to identify the effects of motor placement, response mechanism (e.g., pressing the push-button with hands or legs), or other test conditions.
Figure 4. Typical Calibration Curves. The results of the Fast Fourier Transform of the acceleration data are shown for a single motor undergoing calibration. Four trials were conducted to identify the PWM level corresponding to 180 Hz vibration (solid blue line). Note that vibration varies between the different frequencies. Please click here to view a larger version of this figure.
Figure 5. Reaction Time Representative Results. (A) Response times for individual frequencies are shown. The line-connected data shows the data for able-bodied participants (mean ± SD), while the individual data points represent the three individuals with transfemoral amputations. Reaction times significantly decreased frequency. '*' denotes a significant difference from the 140 Hz reaction times, and the '#' a difference from the 180 Hz frequency, both at significance p<0.05. (B) Response times for distinguishing between pairs of frequencies are plotted for both able-bodied individuals and those with transfemoral amputations. Note that the spread of the data at each pair is much larger than that for the single-frequency data, indicating more variable results. This figure has been modified from data originally published by Sharma et al.10. Please click here to view a larger version of this figure.
PWM Value (pulses) | 64 | 127 | 191 | 255 |
Duty Cycle (%) | 25 | 50 | 75 | 100 |
Table 1. PWM Values and Corresponding Duty Cycles. Sample PWM levels and the corresponding duty cycles are shown. The 0-255 range for the PWM value specifies the number of bytes in each pulse (out of 255 possible) for which the signal is on.
The purpose of this protocol is to provide the framework for evaluating stimulation parameters in vibrotactile ASF applications. Specifically, it examines the effects of vibration frequency, amplitude, location, and sequence on user sensorimotor response. This framework can be built upon and expanded to incorporate additional or alternative types of user response that may be more clinically relevant, such as bending a joint or shifting weight from one leg to another. These types of changes would require slightly different hardware configurations, namely replacement of the push buttons with devices such as inertial measurement units (IMUs) or pressure sensors, as well as the accompanying changes to the virtual interface. Similarly, although the protocol presented here requires the participant to be in a seated position, only small hardware modifications would be necessary to make the transition to more clinically relevant postures, such as standing balance or walking trials.
In both experiments, the push-button(s) may be pressed with the hand, leg, foot, or by some other means, depending on the specific research question and desired response. Furthermore, additional studies employing this basic protocol can be used to explore the effects of different feedback coding strategies, locations, and incorporation into new or existing prostheses. For example, when implementing vibrotactile feedback into lower-limb prostheses, it may be interesting to examine the effects of the prosthetic socket and liner on user sensitivity to the stimuli. While the protocols detailed in this manuscript require semi-manual activation of the vibratory motors (through the interface), they can easily be modified to enable motor activation in response to kinetic or kinematic measurements from external sensors. Using measurement devices, such as IMUs, goniometers, pressures sensors, etc., in lieu of the push button, experiments can be conducted to examine more physiologically relevant user response times and accuracy to vibrotactile feedback provided. This type of study would employ a similar protocols to those described in Experiments 1 and 2, but would require an additional control system to convert the sensor input into instructions sent to the vibratory motors, as well as changes to the data acquisition software to interface with the new hardware changes.
One example of implementing a physiologically relevant response is to replace the push-button with a goniometer to measure changes in knee angle. For this type of experiment, the goniometer would be mounted on the lateral surface of the knee joint, and rather than pressing the push button, participants would be instructed to bend their knee to a pre-defined knee angle (e.g., 90 degrees) upon perception of a motor vibration. User reaction times are then defined as the time elapsed between applied stimulus and when the joint angle settles on or near the desired value (e.g., 90° ± 10°). Movement accuracy can also be evaluated by calculating the percent error between the target and achieved angles.
Over the past ten years, a number of studies have explored the use of vibrotactile feedback in a variety of biomechanical applications, including its efficacy as a training tool for improving gait and balance14,15. The majority of these studies have focused on the clinical implications of biofeedback, examining any changes to specific kinetic or kinematic parameters when vibrotactile feedback is applied. As such, most protocols select a single set of stimulation parameters, with few examining user sensitivity to vibration location, amplitude, or pattern. The protocol presented here serves as a first-step towards understanding user perception to vibrotactile stimuli that should be performed prior to evaluating the effects of these stimuli on specific clinical conditions. Additional work, such as that by Goodworth et al.7,16, which explored various coding strategies for translating sensory information into vibratory stimuli, and Crea et al.12, which evaluated user sensitivity to changes in vibration patterns, complement these experiments to provide a more complete understanding of how to optimize vibrotactile feedback for specific biomechanics applications.
It should be noted that a fundamental limitation of this experimental system, as with other systems reported in the literature5,6, lies with the vibratory motors, which couple vibration frequency and magnitude. That is, increases or decreases in vibration frequency are accompanied by proportional changes in amplitude. Separation of these two parameters requires a different type of motor, such as linear resonance actuators, as well as more advanced motor drivers to power the more sophisticated motors. Additional updates to the existing interface would be required to accommodate the new hardware and additional amplitude parameter.
The calibration procedure is critical to the successful execution of these experiments, and should be performed independently for each motor used in the subsequent experiments. While the duty cycle-frequency relationship should be nominally the same type for identical motors, small differences in motor construction may result in non-trivial changes to resulting frequencies. For example, while developing this procedure, a 180 Hz target frequency was achieved using PWM values ranging from 103-143 for different motors.
The authors have nothing to disclose.
This protocol was developed for research supported by the Natural Sciences and Engineering Research Council of Canada (grant RGPIN 401963).
Vibrating Pager Motors | Precision Microdrives | Model 310-101 | Coin eccentric rotating mass motors. As many as necessary to test all locations and interactions of interest |
Tri-axis Accelerometer | Dimension Engineering | ADXL 335 | Advanced analog accelerometer. 500Hz bandwidth, 3.5-15V input. Designed for motion, tilt, and slope measurement, as well as vibration and shock sensing |
Arduino Uno | Arduino | DEV-11021 | Microcontroller board for communicating with the tri-axis accelerometer |
Arduion Mega 2560 | Arduino | DEV-11061 | Microcontroller board for interfacing with the vibration motors. |
LabVIEW | National Instruments | Data acquisition software used to control motors and display accelerometer signals | |
Arduino IDE Software | Arduino | v. 1.6.5 | |
Push-Button | Bridges | Buddy Button | Wired switch featuring a 2.5in/6.35cm activation surface that provides an auditory click and tactile feedback. |
Optional: | |||
Dedicated haptic motor driver | Texas Instruments | DRV2605L | Can be used to replace the entire amplification circuit described in Step 1. |
Flexible wearable goniometer | Biometrics Ltd. | SG110 | Twin axis flexible goniometers to measure angles in up to two planes of movement that can be used in lieu of the push button to measure joint movement in response to stimuli. www.biometricsltd.com/gonio.htm |