Athena Project
Jaime Ciriaco
Michael Dunn
Aaron Marquez
Sonoma State University Department of Engineering Science
Advisor: Farid Farahmand
Client: Arthur Obuchowicz
http://athena490.wordpress.com
Introduction
• One of the most common disorders in the world is epilepsy
• It can be dangerous for someone with epilepsy to be left home alone because a seizure can leave them
incapacitated and unable to call for help
• This can be especially dangerous if certain appliances such as a stove or iron were left on
• There are only a few devices that monitor this disorder
• Efficient seizure detection algorithms are needed
• If a person has a seizure then they could be incapacitated for a period of time.
Hardware
• 9 Degree of Freedom Sensor: Accelerometer, Gyroscope, Magnetometer
• Feather M0 microcontroller
• Buzzer
Accelerometer/Gyroscope Data
• Accelerometer and Gyroscope data used for analysis
• Velocity (m/s^2) in x, y, z axes
• Rotational Velocity (degrees-per-second) in
x, y, z axes
• 20 Hz Sample Rate New sample every
50 ms
Accelerometer/Gyroscope Data
• Accelerometer and Gyroscope data used for analysis
VX(n) Vy(n), Vz(n)
VX(n) Vy(n), Vz(n), fx, fy, fz
Received Accelerometer Data
Seizure Event Random Movement
Vx(n) Vy(n), Vz(n)
Received Accelerometer Data
Seizure Event Random Movement
Vx(n) Vy(n), Vz(n)
Problem: How do you
detect Seizure?
Introducing Root-Mean-Square Analysis
• Read accelerometer & gyroscopic values
• Calculate RMS
V RMS =Vx
2 +Vy2 +Vz
2
3
RMS =1
Nvi
2
i
N
åGeneral Equation:
Our Implementation:
Seizure Detection Algorithm
• Calculate Difference: VRMS(n) – VRMS(n-1) > THRMS
• Calculate Counter: Count(n) = Count(n-1) + 1 > THC
Necessary Model
• Source: T. R. Burchfield and S. Venkatesan,"Accelerometer-Based Human
Abnormal Movement Detection in Wireless Sensor Networks”
THRMS
Our Model
• Green Arrow (THRMS): “Shaking event”
• Red Line (THC): “Possible Seizure Event”
>THRMS >THC
Including Fast Fourier Transform
in Seizure Detection Algorithm • Look at RMS data in frequency domain
X(k) = x( j)wN
( j-1)(k-1)
j=1
N
å
where
wN = e(-2pi)/N
FFTRMS (k) = VRMS ( j)wN
( j-1)(k-1)
j=1
N
å
K=Number of Samples
Seizure Detection Algorithm with FFT
If certain frequency found AND THC passed Alarm
FFT of RMS (VX, Vy, Vz) Data
FFTRMS (k) = VRMS ( j)wN
( j-1)(k-1)
j=1
N
å
FFT of Individual VX, Vy, Vz Data
FFT (k) = VX,Y ,Z ( j)wN
( j-1)(k-1)
j=1
N
å
FFT of Gyroscope (fx, fy, fz) Data
FFTx,y,z (k) = Fx, y, z( j)wN
( j-1)(k-1)
j=1
N
å
Final Seizure Algorithm
Conclusion
• Accuracy of model
• Speed of Model