rocbuilder

[rocbuilder]

Usb gps

Got a usb gps, because the previous one got damaged

https://www.raspberrypi.org/forums/viewtopic.php?t=56023

I still haven't figured out how to use python to get the gps data

[rocbuilder]

Raspberry pi keyboard

After plugging in a new keyboard, i remembered there was a strange issue, being that the raspberrry pi doesn't properly recognise the keyboard. The \ key returns a # for example

https://www.raspberrypi.org/forums/viewtopic.php?f=26&t=39806&p=331516#p331516

This was my reference to solving that problem

[rocbuilder]

How to gui?

Because i have little idea how to make a gui for python, i searched the webs and came across this https://www.daniweb.com/programming/software-development/code/216785/tkinter-digital-clock-python It's code for a clock, and i used this clock code to display the strokerate i find

[rocbuilder]

Reflection of project thus far

So this is the second time i’m attempting this project to make a strokerate counter + gps with a large display screen

The thing about this second attempt is that i was trying out new methods in a failed attempt to make a more efficient code than the first

So one of the things i obsessed over was frequency analysis, trying to get FFT to work, because i assumed it would be an effective way for the raspberry pi to find the frequency compared to what i’d been doing prev which was to find peaks and calculate the freq from there.

The problem with the previous method was that i had many false positives, and the strokerate calculated would be inaccurate and imprecise

The method using fft was above my understanding and resulted in a lot of time trying to decipher syntax

After a couple weeks of that train wreck, i realised i could attempt to improve the previous algorithm, and very quickly, like within a night, produced a new algorithm that worked, having higher accuracy and precision, as well as a faster read time, also being a much shorter length of code

I really banged my head on the table when i realised those simple improvements

[rocbuilder]

A brief summary thus far

This'll be a summary of what the code does 1. Every 0.05 seconds: I. get acceleration data from sensor II. Calibrate acceleration data III. Apply low pass filter IV. Get summation of acceleration vectors V. If zero crossing, append time to list 2. Every 1 second I. Calculate frequency from 1.V II. Output strokes-per-minute 3. Every 1 minute or longer I. Find mean of each acceleration axis from past 10 seconds

[rocbuilder]

Waveform origin

Because each axis of the accelerometer's origin points defer depending on the orientation of the sensor (because gravity), they need to be set to zero So a "base" value needs to be found for each axis. To find this value, simply have a function that runs either every few minutes, or when it is detected that the boat is not doing any stroking. Just record the values across the past 10 seconds or so, and get the mean The values to be analysed would be the raw data value - the base value

[rocbuilder]

Frequency estimation

Instead of looking at it as peak detection, it makes more sense to look at it is frequency estimation. So how to find the frequency? Previously it was suggested to do peak detection, but the problem with this is, there occasionally can be double peaks during the same stroke, also a threshold value needs to be implemented such that noise is ignored Because the number of peaks during a single stroke might not be consistent, a new method was quickly found https://gist.github.com/endolith/255291 Because the zero crossing is definite between strokes, it is a much more sound solution. And the noise can similarly be ignored by using a threshold value A bit more thought would have to be put to this before something definite arises

[rocbuilder]

Peak detection

How to detect peaks? The sensor values might vary wildly, a threshold can be implemented such that values below a certain point are ignored Then if a certain value is more than the one before it and the one after it, it's a peak Or maybe to add on, peak if (y(t) - y(t-dt) > m) && (y(t) - y(t+dt) > m) - http://stackoverflow.com/a/22583761

[rocbuilder]

Accelerometer algo

Okay, so let me get this straight, i’ve been thinking about this for 2 weeks and barely have a conclusion

So how we get stroke rate is by measuring acceleration by use of a 3 axis analog accelerometer. The accelerometer being analog and all, gives a waveform where the origin point, or long average, is more than zero. In addition, the values might be choppy, large amounts of inaccuracy.

problem 1: choppy values. Solution: implement an smoothing function

Problem 2: origin not 0 Solution: subtract long average

Problem 3: 3-axis Solution: summation by Pythagoras theorem

But now arises a new problem where Pythagoras theorem always produces a positive value

Problem 4: pythagoras theorem always positive Solution: find axis with largest value, if axis more than origin, result is positive, if less than, result negative

So after all that, the waveform is something useable. Next is to read it, find the peaks…

[rocbuilder]

Gps module and uart

The gps module i procured uses the uart protocol It took me a very long time to figure out how to read and use the data from a gps through UART on the arduino, and there only seems to be less documentation for the raspberry pi, godspeed There was a post that suggested to connect the gps module through usb instead, but that'd require an adapter, which maybe can be crudely fashioned through a spare usb cable Also, there's this guide that goes through how to read/write from the serial port http://www.elinux.org/Serial_port_programming And the following step would be to read the nmea gps protocol to spit out the relevant data The worst part about using the gps is that my home is pretty much a faraday cage with weak reception for some 10 mins walk

[rocbuilder]

Python multi-threading

With c++, when i want to run a task ever N seconds, i'd use the delay function or millis function if there are multiple tasks. Thankfully, python comes with a very easy to use multithreading library http://stackoverflow.com/questions/3393612/run-certain-code-every-n-seconds

[rocbuilder]

Raspberry pi 2, wifi

Occasionally, i have to download libraries or upload code to the cloud, all this requires internet access. Unfortunately, the raspberry pi doesn't have wifi capabilities, and requires an ethernet cable. Becomes inconvenient especially since my desktop isn't near an ethernet port So i bought a wifi dongle, and i'll be following this guide to hopefully get wifi capabilities http://raspberrypihq.com/how-to-add-wifi-to-the-raspberry-pi/

[rocbuilder]

Storing data

http://www.pythonforbeginners.com/systems-programming/using-the-csv-module-in-python/

by using csv, the large amounts of data from the sensors can be read and written easily

[rocbuilder]

Visual analysis

matplotlib is able to provide some tools for analysing the sensor data

this here post gives some examples of guis to go along with matplotlib

[rocbuilder]

saving files in python

I referred to this post regarding how to save files to a specific directory

Then using the date and time, a unique name can be made for each file.

And lastly, an infinite loop can be used to keep writing data to the file

[rocbuilder]

Analog to digital

The raspberry pi does not have analog gpio, however, by using an analogue to digital converter (ADC), analog data can be read.

Thankfully, documentation on how to do so has already been provided https://www.raspberrypi.org/learning/physical-computing-with-python/analogue/

[rocbuilder]

"Pisces" : A Raspberry Pi Powered Boat Nav/GPS with Topographic Map Viewer

https://www.reddit.com/r/raspberry_pi/comments/65ouwu/if_it_worked_any_better_it_would_have_put_the/?st=J1Q2JXUD&sh=224e6572 This is very relevant to the project i'm doing The creator mentioned to get to know glade and gtx, so i'll have to look into those. I suppose the first iteration of my gui would look similar to that