1. Abstract The investigation of proposed auditory-based
neurofeedback of EMG signals from leg muscles of walking
human volunteers for neuromuscular rehabilitation is presented.
Off-line simulations based on collected EMG signals resulted in
audio signals in which phases of the step cycle could be
discriminated by ear. The method will be implemented as a
portable device for more personal healthcare exercises.
I. INTRODUCTION
As a result of various pathologies (TBI, stroke, spinal cord
injury, neurodegenerative pathologies), patients experience
altered gait as a major impediment to activities of daily living.
Physical therapy for gait rehabilitation/gait training in
individuals with stride abnormalities typically involves
professional therapy services in order to normalize their gait
patterns, and established protocols for rehabilitating
individuals with gait abnormalities typically involve
participation in a therapy facility for periods of 2-4 months.
While these protocols clearly function, patients tend to incur
substantial cost and time commitment. It has been shown that
biofeedback of step-cycle could provide significant
improvement of walking rehabilitation [1]; therefore, we
proposed a gait rehabilitation therapy device using auditory
feedback based on electromyogram (EMG) signals from the
muscles of interest for more personal therapy.
II. TECHNIQUES
MATLAB simulations (MATLAB 8.5, Mathworks, Inc)
were conducted offline using previously acquired EMG data.
Surface electromyogram recordings of quadriceps, hamstring,
and calf muscles were acquired using Trigno EMG sensors
(Delsys, Inc) at a sampling rate of 2 kHz while the user
performed normal walking. Muscle signals were first highpass
filtered at 0.5Hz to remove bias signal and then digitally
converted and stored. The EMG signals were then rectified and
low-pass filtered at 4Hz in MATLAB to give signals
representing the overall muscle activity for that muscle group.
These signals were then normalized and used to modulate the
amplitude of three tones (quadriceps 100 Hz tone,
hamstrings 150 Hz tone, calf 200 Hz). Tones generated by
the muscles of the left and right leg were combined into
separate mono audio files, and then combined into a final
stereo audio file.
III. RESULTS AND FUTURE DIRECTION
The offline gait sonification resulted in an audio signal that
allowed auditory discrimination of gait patterns. The raw
EMG signals (Figure 1A) show the periodic contractions of
the leg muscles during step cycles, and these periodicities are
reflected in the corresponding audio signal (shown as a
spectrogram in Figure 1B).
A system is currently being developed for implementing
this sonification process in a mobile, real-time device. This
system consists of EMG circuitry for signal acquisition, and a
microcontroller (TM4CG123, Texas Instruments) for
performing sonification algorithms and sending real-time
MIDI signals to a MIDI decoder (VS1053, VLSI Solution)
connected to a pair of speakers. The flexibility of the MIDI
protocol allows for testing a variety of auditory feedback
paradigms, including the amplitude modulation paradigm
shown in the above simulations.
Figure 1. A) The
EMG signal during
normal walking.
Top (Red) signal
represents
Quadriceps, Middle
(Green) signal
represents
Hamstring, and
Bottom (Blue)
signal represents
Calf. All units in
millivolts.
B) The sonified
signal. 100Hz
represents
Quadriceps, 150Hz
represents
hamstrings, and
200Hz represents
calf. Color scaled
for power, with
brighter (Yellow)
indicating higher
power and darker
(Blue) indicating
lower power. No
noise was generated
outside of the
spectrum range.
REFERENCES
[1] D. Intiso, V. Santilli, M. G. Grasso, R. Rossi, and I. Caruso,
foot- Stroke, vol. 25, no. 6, pp. 1189 1192, Jun.
1994.
P. Donaldson, Member, IEEE, A. Gopinath, D. Zuniga, and J. Cagle, Member, IEEE
*Research supported by University of Florida Biomedical Engineering
Society Student Chapter Research and Development Team. P. Donaldson is
a student with the J. Crayton Pruitt Family Department of Biomedical
Engineering and Department of Electrical and Computer Engineering, and J.
Cagle is the Research and Development Team Lead and a student with the J.
Crayton Pruitt Family Department of Biomedical Engineering, University of
Florida, Gainesville, Florida 32601 USA.