Date of Award

12-15-2014

Document Type

Open Access Dissertation

Department

Electrical Engineering

First Advisor

Yong-June Shin

Abstract

Falls in the senior population represent an immediate threat to both current and future generations’ quality of life. The results of falls can be disastrous, create a long road to recovery, and in many cases result in death. Hypertension represents a difficult-to-quantify condition which is known to contribute to gait and balance dysfunction. Accurate assessment of conditions such as these represents an area of primary investigative need. In order to meet this need, a new experimental setup is developed which combines time-frequency analysis with surface electromyography (sEMG) signals obtained during ambulation. 3-bit pressure data is acquired using a pressure-sensitive mat, which records time-aligned gait information alongside 8 dipole sEMG sensors attached to tibialis anterior, gastrocnemius, biceps femoris and vastus lateralis. These muscles serve as representative muscles for the ankle flexor, calf (knee flexor), hamstring and quadriceps groups. In all muscles save the biceps femoris, these represent largest muscle in each respective group, providing a comprehensive picture of muscular activation during gait. The sEMG signals are recorded simultaneously with those of a pressure-sensitive mat, the data from which is used to identify the sinusoidal center of mass for gait separation. The two signals are time-aligned using a recording trigger sent from a standalone digital output device. Signals are imported into MATLAB and rejected according to Grubb's test using equal ninety percent on each of three signal attributes: signal energy, signal dc level, signal and peak-to-peak voltage. Signals are decomposed using a Hann window reduced interference distribution, with an Nt = 53-point Hann time window and N! = 96-point Blackmann frequency window. Six metrics based on the physiology and spectral analysis of sEMG signals are used to evaluate and compare several population groups. These metrics are: Instantaneous time duration(TD%), local frequency bandwidth (FB), local frequency maximum (Fmax), energy ratio conditional (E!b ), conditional energy from 40-100 Hz (E%40−100Hz), and energy spectral density from 40-100 Hz (ESD40−100Hz). These metrics rely on both Fourier transform spectrum distributions and accurate timefrequency localized distributions. Using a large subject database, 10 male and 20 female controls are compared to 7 male and 17 female hypertensive subjects across the six metrics. Subjects are separated into control and experimental groups using medical history and self-report data; age was not a factor. Each of the six metrics are then evaluated in the four muscle groups for differentiability between control and experimental groups. These metrics are then evaluated in male and female subgroups. In the male subgroup, when using the tibialis anterior muscle, TD%, FB, and E!b=120Hz showed a maximal accuracy of 75.00%, 76.92% and 76.92% respectively. In the gastrocnemius, both FB and E!b=120Hz showed 76.92% accuracy. In the vastus lateralis, Fmax), and E!b=120Hz showed 76.92% accuracy, while E!b=40Hz showed 75.00% accuracy. The biceps femoris showed low levels of accuracy (maximum 62.50%). In the female subgroup, the overall level of accuracy is lower, due to physiological factors of subcutaneous tissue, muscle distribution, and gait differences. In the tibialis anterior, E!b=40Hz showed the highest accuracy at 68.42%, with TD% at 66.67%. In the gastrocnemius, TD% showed highest accuracy. In the vastus lateralis, E!b=120Hz showed the highest accuracy, and in the biceps femoris, E!b=40Hz showed the highest level of accuracy. Using an aggregate of these key metrics, an accuracy of 94.12% for male subsets and 78.38% for female subsets is established for testing control groups vs hypertensive experimental groups. This research represents a hypertension diagnostic tool, and thus a quantitative indicator of fall risk.

Share

COinS