First of all, I would like to thank the Sally McDonnell Barksdale Honors College and Department of Electrical Engineering for providing me with all the amazing courses and massive resources that help me with my work. James Sabatier, the managing member of SOAIR LLC for supporting me with the research and performing all the experiments. I thank my colleagues who worked with me and all the other SOAIR employees and former employees on the same project for help.
Last but not least, I would like to thank my family for respecting my choices in the academic profession and for all my support. A fourth-generation gait measurement device is designed to capture and analyze detailed gait and step metrics, which ultimately provide a Fall Risk Assessment score. Specifically, the device is modified to suit the residential environment and elderly consumers, which is low-cost, user-friendly and portable.
Overall, the device has been built and shown to have better performance than previous generations. The built-in gait analysis program ran slower than the computer version program, but has the same accuracy.
INTRODUCTION
BACKGROUND
PREVIOUS WORK AND ACHIEVEMENTS
The second-generation prototype contained a preamplifier, ultrasonic transducers, and a data acquisition device (NI-USB 6216 DAC) in a box that is portable for field testing. The box is redesigned with different components that have the same functions as each other to achieve low cost.
CHALLENGES
DESIGN OF THE WALKING GAIT MEASURMENT DEVICE
- SYSTEM LAYOUT
- MAIN OPERATION AND FUNCTION DESCRIPTION
- OPERATION PRINCIPLE
- ULTRASONIC SIGNAL TRANSMISSION AND RECEIVING
- ANALOG TO DIGITAL (A/D) CONVERSION AND
- SIGNAL PROCESSING
- OTHER FUNCTIONS
- IMPROVEMENTS
- PASSIVE INFRARED (PIR) SENSOR
- ANALOG TO DIGITAL CONVERSION
As shown in Figure 2.1, the device operated with two ultrasonic transducers, one as a transmitter and the other as a receiver. The receiver is turned on at the same time to receive the signal from the vibrating surface due to the Doppler effect [16]. Moving objects cause a phase shift in the signal used to detect the motion.
As long as the raspberry pi 3 receives the demodulated data, an on-board image processing program extracts the gait parameters, the detail of which will be discussed in Section 3. However, MATLAB does not have a raspberry pi version library or library written for ARM-based CPU. The device can either be automatically activated by a passive infrared (PIR) sensor (on PIR mode) or manually activated by another raspberry pi with touch screen (on Bluetooth mode).
After it is activated, the device will take data for ten seconds and temporarily store it in the in-box raspberry pi. Since the passage information can be accessed by the onboard programs, it can be uploaded to the online database or sent to any connected personal devices. Compared to the 3rd generation, the 4th generation gait measuring device improves its performance in several aspects.
Second, the improvement of the PIR sensor reduces data redundancy and data verification which will be discussed in Section 2.3. The PIR sensor in the 3rd generation device has a wider detection range than the ultrasonic transducer. This means that the device will be falsely activated even the target is out of the effective detection range.
The solution is to use a Fresnel lens to narrow the detection range of the PIR sensor. Converged by the customized Fresnel lens EWA 0.4 GI V1, the analog version PIR sensor LHI986 comes out with a narrow detection width on one side and a wide area when rotated 90 degrees. On the other hand, since the detection area is wide, the movement of people with different heights can also be easily detected.
In the third generation device, the ADC (LTI Board) was built using the I2S protocol which can only operate at multiples of 24 kHz, such as 96 and 192 kHz. Considering that the algorithm standard has been established and tested throughout the previous generations of the device, the first attempt is to work out another ADC solution to match the preferred parameter.
GAIT ANALYSIS
STRIDE DATA EXTRATION
To extract the step pattern, the spectrogram is first converted to gray code where a threshold is set, which improves the contrast. Then a Wiener filter is applied to minimize the mean squared error between the estimated random process and the desired process, in other words, to reduce the Gaussian noise [18]. During walking, toe and heel have the maximum velocity represented by the envelope of the spectrogram.
To fit the curve to the envelope data and find the expression, the envelope data is smoothed as shown in Figure 3.4. Then, fit the speed on the log scale with multiple parabolas that have the maximum speed noted. Finally, select the coefficients for the Gaussian function to obtain the original expressions for the curve.
The purpose of this step is to remove the scattered parts near the bottom of the Gaussian functions, since there is a time when the foot velocity approaches zero, recognized as the steady state time. The location of each cosine function is determined by the peak of each wave found in the previous steps.
GAIT DATA CALCULATION AND APPROXIMATION
- PEAK FOOT VELOCITY
- STANCE TIME
- SWING TIME
- STRIDE TIME (GAIT CYCLE)
- SWING/STANCE TIME
- STANCE TIME PERCENTAGE
- STEP CADENCE
- STRIDE DISTANCE (STRIDE LENGTH)
- AVERAGE STRIDE VELOCITY
- DOUBLE STANCE TIME
- SYMMETRY INDEX
- GAIT SPEED
- WALK RATIO
- RESULTS EXAMPLE
The stance time is the duration from ipsilateral foot contact to ipsilateral foot off, which is expressed as "iTO β iIC1" shown in Figure 3.6. The average stance time in the gait analysis procedure represents the average stance time of both feet. Swing time is the duration from ipsilateral foot off to ipsilateral foot contact or the time the foot is in the air.
The average swing time in the gait analysis procedure represents the average swing time of both feet. A single step time is expressed as "iIC2 β iIC1" shown in Figure 3.6 or simply "dwell time". The average walking time in the gait analysis procedure represents the average walking time of both feet.
As it is mentioned, the swing-to-stance ratio is equal to the swing time divided by the stance time. The average swing/stance ratio in the gait analysis procedure represents the average of each ratio of both feet. The average stance time percentage in the gait analysis procedure represents the average of each percentage of both feet.
Average stride speed, also known as average foot speed, is defined as the average speed of the foot during the swing. The mean step speed mean is the average of the step speed of all steps. Double stance time is the time when both feet touch the ground, from the heel down on one foot to the toes on the other [23].
An initial double stance time is expressed as βcTO β iIC1β and a late double stance time is expressed as βiTO β cICβ, depending on which foot is defined as the ipsilateral one. The average double stance time is the average of all double stance times during a walk. The mean value of the walking speed in the procedure represents the average walking speed for each step.
CONCLUSION AND FUTURE DEVELOPMENT
CONCLUSION
FUTURE DEVELOPMENT
Verghese et al, βQuantitative gait dysfunction and risk of cognitive decline and dementia,β Journal of Neurology, Neurosurgery, and Psychiatry, vol. Taylor et al, βGait parameter risk factors for falls under single- and dual-task conditions in older people with cognitive impairment,β Gait and Posture, vol. Park et al, "Simulation and analysis of polarimetric radar signatures of human walking," IEEE Transactions on Aerospace and Electronic Systems, vol.
Xue et al, βInfrared gait recognition based on wavelet transform and support vector machine,β Pattern Recognition, vol. MatiΔ et al, βInfrared assessment of knee instability in patients with ACL deficiency,β International Orthopaedics, vol. Ekimov, "Rhythm analysis of orthogonal signals from human walking," The Journal of the Acoustic Society of America, vol.
Komma, "Ultrasonic Doppler Sonar Sensor System for Extraction of Walking Signature", Undergraduate, University of Mississippi, 2016. Bradley et al, "Extraction of the velocity of walking human body segments using ultrasonic Doppler," The Journal of the Acoustical Society of America, vol. Damarla, βUltrasonic Doppler Methods for Extracting Signatures of a Walking Human,β The Journal of the Acoustical Society of America, vol.
Buckley, βThe effects of walking speed on minimum toe play and on the temporal relationship between minimum toe play and maximum swing foot speed in unilateral trans-tibial amputees,β Prosthetics and Orthotics International, vol. Beauchet et al., βWalking speed-related changes in stride time variability: effects of reduced speed,β Journal of Neuroengineering and Rehabilitation, vol. Iosa et al., βThe golden ratio of gait harmony: repetitive proportions of repetitive gait phases,β BioMed Research International, vol.
Loke, βSwing Phase Control and Stance Phase Control,β Swing Phase Control - Stance Phase Control - DBSβ’.
