<solved> Rollover Handling for R-R Intervals via Page 4

Username		Password		Remember Me?	forgot password?

20 October 2021 07:30 PM

Total Posts: 12

Joined 2020-02-12

First, thanks in advance! I need some feedback on my script. I'm using python 3.7 to capture my various device data streams for later analysis. That later portion will eventually deal with DFA-a1 analysis using this data.

Here's the code portion that calculates the R-R interval based on Package (Page 4) as noted on Section 7.2, page 33 of the HR device profile (file excerpt attached to this message):

# This section is for the POLAR sensor, deviceNumber=18990
if msg.deviceType == 120:
    mv = memoryview(msg._content)
    # Page 0x04. Use the recommended main page 4 only
    if mv[0] == 4:      
        # Assembled computation elements (LSB/MSB)
        val_HR = mv[7]  # Calculated HR (bpm)
        val_HRV = abs( ((mv[5]<<8 | mv[4]) - (mv[3]<<8 | mv[2])) *1000/1024 )  # HRV (milliseconds) 

The resulting R-R interval value (val_HRV) seems reduced a bit when compared to a concurrently running HRV reader app. using the same Polar H10 strap as my script above... close, but not a match. They should be the same correct? This brings me to the next part, rollover handling. I'm certain I don't have the intent coded correctly to address rollover for the heartbeat count and the heartbeat event time. I'm not certain what to do with the differences. Here's what the code portion looks like:

# Handle rollover events
self.heartbeatCount = mv[6]
self.heartbeatEventTime = val_HRV

if self.heartbeatCountOld > self.heartbeatCount:
    # The rollover correction at 2^8
    heartbeatCountDiff = (self.heartbeatCount + 256) - self.heartbeatCountOld
else:
    heartbeatCountDiff = self.heartbeatCount - self.heartbeatCountOld

if self.heartbeatEventTimeOld > self.heartbeatEventTime:
    # Rollover correction at 2^16 = 65536, or 63.999 * 1024 either is about 65,534.976
    heartbeatEventTimeDiff = (self.heartbeatEventTime + 65536) - self.heartbeatEventTimeOld
else:
    heartbeatEventTimeDiff = self.heartbeatEventTime - self.heartbeatEventTimeOld 

Also, I'm not certain that the "else" conditions belong as Section 7.1 doesn't mention it. I sure would appreciate where I'm going off-track as I don't want to get into the DFAa1 phase with inaccurate data.

What to do with the rollover adjustments/corrections?

TIA,
Eric

Image Attachments

Click thumbnail to see full-size image

25 October 2021 10:19 AM #1

haighi Administrator

Total Posts: 370

Joined 2012-06-27

PM

"val_HRV = abs( ((mv[5]<<8 | mv[4]) - (mv[3]<<8 | mv[2])) *1000/1024 )" does not correctly handle rollover in python. Ex abs((10 - 65535)*1000/1024) = abs(-63989.26) = 63989.26. It should be 11*1000/1024. In embedded c with unsigned types the rollover is handled by the nature of unsigned math.

The hr timestamp values rollover at 65535 to 65536 transition. They got from 65535 to 0. This maps to 2^16 in sec 7.1. This means you should be using

"if self.heartbeatEventTimeOld > self.heartbeatEventTime:
# Rollover correction at 2^16 = 65536, or 63.999 * 1024 either is about 65,534.976
heartbeatEventTimeDiff = (self.heartbeatEventTime + 65536) - self.heartbeatEventTimeOld
else:
heartbeatEventTimeDiff = self.heartbeatEventTime - self.heartbeatEventTimeOld "

Signature

Ian Haigh

25 October 2021 10:22 AM #2

haighi Administrator

Total Posts: 370

Joined 2012-06-27

PM

Note: The app you are comparing against may be using section 7.3 and therefore not match if you only use 7.2. You would have to reach out to the app creator to determine what is causing the difference.

Signature

Ian Haigh

25 October 2021 10:25 AM #3

skipper08023002 Newbie

Total Posts: 12

Joined 2020-02-12

PM

Thanks so much! I'll check it out when I get home tonight! Best!
Eric A.

27 October 2021 02:54 AM #4

skipper08023002 Newbie

Total Posts: 12

Joined 2020-02-12

PM

Ian,
That other app creator's work is based on Polar's BLE implementation, so I'd expect the output values wouldn't be an apples-to-apples comparison.

I re-thought the integration of the rollover condition in the above script, changing it to:

# This section is for the POLAR sensor, deviceNumber=18990
if msg.deviceType == 120:
    mv = memoryview(msg._content)
    # Page 0x04. Use the recommended main Page 4
    if mv[0] == 4:       
        # Assembled computation elements (LSB/MSB)
        val_HR = mv[7]  # Provided HR (bpm)
        # Initiate event values
        self.heartbeatEventTime = (mv[5]<<8 | mv[4])
        self.heartbeatEventTimeOld = (mv[3]<<8 | mv[2])
        # Check for rollover by comparing new vs. old values
        if self.heartbeatEventTime < self.heartbeatEventTimeOld:
            # Rollover has occurred. Correction at 2^16 = 65536
            val_HRV = ((self.heartbeatEventTime + 65536) - self.heartbeatEventTimeOld )*1000/1024
        else:
            val_HRV = (self.heartbeatEventTime - self.heartbeatEventTimeOld)*1000/1024 

I haven't found another reader/tracker app that uses the ANT+ Device Profile (Section 7.2 specifically) but will continue to search. Thoughts on the above?
TIA,
Eric

02 November 2021 09:55 AM #5

haighi Administrator

Total Posts: 370

Joined 2012-06-27

PM

The above looks correct. SimulANT+ parses R-R information.

Signature

Ian Haigh

02 November 2021 09:59 AM #6

skipper08023002 Newbie

Total Posts: 12

Joined 2020-02-12

PM

Ahh, very good. Many thanks, Ian.
Cheers,
Eric A.

ANT Forum