adammansfield / real-time-heart-rate-detection

An ultra low power, real-time heart rate detector on the MSP430.

RealTimeHeartrateDetection

An ultra low power, real-time heart rate detector (QRS detection) on the MSP430F5529.

Part List

• Ultra Low Power Mircontroller: Texas Instruments MSP430F5529
• Raw LCD: Varitronix 3-Digit Reflective LCD
• ECG Circuit: Instrumentation amplifier, op-amps, resistors, capacitors, and leads

Dependencies

• Code Composer Studio v5.5.0.00077+

Power Consumption

In active mode:

Component	Current	Power @ 3V
MCU	290 μA	870 μW
LCD	30 μA	90 μW
Total	320 μA	960 μW

In low power mode:

Component	Current	Power @ 3V
MCU	80 μA	240 μW
LCD	30 μA	90 μW
Total	110 μA	330 μW

QRS Detection Algorithm

A moving-average-based QRS detection algorithm is used to detect heartbeats. This algorithm has three steps: 1) a moving average-based highpass filter, 2) a nonlinear lowpass filter and 3) threshold detection.

Step 1 - High Pass Filter
This step's purpose is to suppress baseline drift and emphasize the QRS complex.

x: input data
M: moving average window size for high-pass
y1: moving average filter
y2: delayed input sample
y: FIR high pass filter with linear phase output

               M-1
y1[n] = 1/M  *  Σ (x[n-m])
               m=0
y2[n] = x[n - (M+1)/2]
y[n] = y2[n] - y1[n]

Step 2 - Low Pass Filter
This step's purpose is to accentuate peaks.

M: moving average window size for low-pass
h: data points from low-pass filter
z: nonlinear lowpass filter output

z_sum[0] = h[0]^2 + h[1]^2 + h[2]^2 + ... + h[M-1]^2
z_sum[1] = h[1]^2 + h[2]^2 + h[3]^2 + ... + h[M]^2
z_sum[2] = h[2]^2 + h[3]^2 + h[4]^2 + ... + h[M+1]^2
 .
 .  i goes to M-1
 .
z_sum[i] = h[i]^2 + h[i+1]^2 + h[i+2]^2 + ... + h[i+M-1]^2

       M-1
z[n] =  Σ z_sum[i]
       i=0

Step 3 - Threshold Detection
This step's purpose is to detect heartbeats. A heartbeat is detected if the peak of a detection frame is above the calculated threshold. The next threshold is then calculated based on the old threshold and the new peak.

alpha: forgetting factor (range from 0.01 to 0.10)
gamma: weighting factor for new_peak (0.15 or 0.20)
new_peak: local maximum detected in new frame

new_threshold = alpha * gamma * new_peak + (1-alpha) * old_threshold

Source: H.C. Chen and S.W. Chen, “A Moving Average based Filtering System with its Application to Real-time QRS Detection,” IEEE Computers in Cardiology, 2003, pp.585-588. Link to PDF

Pin Map from MSP430 to LCD

MSP430	7SEG	LCD
P4.0	COM	24
P3.0	1E	9
P3.1	1D	10
P3.2	1C	11
P3.3	1B	12
P3.4	1A	13
P3.5	1F	14
P3.6	1G	15
P6.0	2E	5
P6.1	2D	6
P6.2	2C	7
P6.3	2B	16
P6.4	2A	17
P6.5	2F	18
P6.6	2G	19
P2.0	3C	3
P2.2	3B	20

Third Party Sources

• consoleio.h/c
  • Original Author: oPossum
  • Source: http://forum.43oh.com/topic/2354-printing-to-the-ccs-debugger-console/
• printf.h/c
  • Original Author: oPossum
  • Source: http://forum.43oh.com/topic/1289-tiny-printf-c-version/