Hi arswad

For sure the algorithm could be improved.
Corelation would require to have a "good" record already.
I worked with FFts (or here better DFTs) at university, so I know a bit about it.
How ever I think it might make the whole application to complicated and battery draining.

I am still surprised that some people can see 90 minutes cycles in their sleep, as for me I usually can not see them. But this might be as I have to sleep in a noisi environment.
In some weeks I will move, then I will see how my sleep pattern changes.

Hi caco3,

It is true that some nights I get very good 90min cycles recorded. However, as you said it depends on too may factors. I sometimes toss and turn closer/further from the phone and that will give very different results.

It most likely would make the app complicated and battery draining, but it would still be a lot better than what I am doing at the moment - using alarmed to trigger a script to play a mp3 file at a preconfiged time. I haven't had any good results with that method since the timing is predetermined. Hope you have the time to look into it.

I'm using the app every night now and going to build up a large sample to see if I can draw anything from it, or just plot something fancy in excel or R :P

Yes caco3, I sure agree its somewhat of an overkill just for that single "wakeup" feature. It was just an idea that popped into my head when I saw my nightly graphs. And maybe, like you say, it might be difficult to track reoccuring patterns with some sort of descent length to be able to match it. On top of that, the user would need to save intervals and stuff to reference file....it would be too complex.

One question though, what is that "value smoothing" more specifically, I would guess it is a moving average. But you make it in two steps?
What is actually going in to the recorded file?

I'm aware these things can be easily misunderstood! ;-)

@rexii23:
If you can find an algorithm,I am willing to see if it can be implemented.
The data i use internally is the same as is in the file, so you have all you need.
Matlab, Octave or R are for sure very usefull for that.
Excell sadly runs faar to quickly onto a wall
If needed I also can provide some of my own records.

@Larswad:
I explained the algorithm in one of the earlier post in this thread.

Hi all

This is just a reminder, PLEASE, if you use the latest Testing/Devel version and you are happy with it, please vote for it under http://maemo.org/packages/package_in...alyser/1.6-53/

Thank you a lot!

In some weeks I hopefully will have time again to continue developing.

Abug: Sometimes Sleepanalyzer draws wrongly the diagram.
I had such a thing today and took a screenshot.

In the attached image (below), see the time-axel having hours 6 and 7 twice. Scrolling horizontally didn't fix the problem.

However, when loading from made CVS file, the graphics are ok and also if PNG image is exported.

btw, the application when exporting PNG image could suggest a filename to have that .png suffix. Now it doesn't.

I'd like to be able to add a note to the record after the recording has been stopped, and why not also when the old one is opened the note could be editable.
For example for writing down what kind of dreams you had that night after you woke up.

Hi zimon

That bug looks very weird. I guess it is very hard to reproduce, as you said yourself that it was correct when you loaded the record and exported it. But when it was wrong, did you try to change the zoom level? I am wondering if that would have fixed the problem.

The png extension makes sense and I might add it in the next version.

About note taking:
You always are able to add/edit a note. Just click on the note pad button you can see on the screen shot.

Hi all

In some weeks I want to continue my work on SleepAnalyser.
I would like to get some feedback together:
- What should be improved?
- Any known bugs?
- How can the GUI be improved?
- Which functions are you missing the most?
- Who is interested to translate it into another language?

Some things I already have on my list:
- Multi language
- If possible better visualisation
- GUI improvements

Please also vote for the current version, I still need one vote
http://maemo.org/packages/package_in...alyser/1.6-53/

Best regards
CaCO3

I think you should store the data without condensing/smoothing it up front. So saving every sensor read,
Just do the condensing.smoothing when drawing the pictures.
That way everyone can experiment with the smoothing and derivations (which then of coarse need to be configurable from the gui).

I played a bit with adding the z component also and found it can make a difference. (I guess the n900 can jump up and down when you move around). So I would vote for also saving the z coordinate to the log file.

When you are concerned about bigger datafiles in that case, you can compensate a bit by removing the spaces in the output, and saving relative values and relative timestamps. (that would easily save around 5 bytes per record).

Bug: Only a set number of snoozes are allowed, after which the alarm will not go off again (even though the snooze button has been pressed). Or perhaps this only happens when you do not do anything at all while the alarm rings, and just let it finish instead. Don't remember, perhaps both.