[SOLVED] Importing raw data - or from text list

[Sprinterdriver]

Hi.

I currently posess a large text file with samples. It basically just holds the sample values for left and right channel - one line typically line would hold this information:

• A signed integer between -32768 and +32768 (left channel)
• a tab separator
• signed integer (right channel)

I can always change the format of the list if neccessary in order to import.

I may be able to export as raw data, but I need to know how to assemble it in order to achieve correct stereo, I write what I guess is right and I hope you can correct me:
* No header at all - the samples starts just from first bit.
* 16-bit signed integer (left channel ?)
* 16-bit signed integer (right channel ?)
* 16-bit signed integer (left channel ?)
* 16-bit signed integer (right channel ?)
........
continues like this.



Thanks in advance

[steve]

The first thing is that "text" is not "raw data".
"Raw data" is a series of binary numbers, 0's and 1's.
"Text" is a series of "characters". Each character is represented in a file or data stream by a binary number. The relationship between the number and the character depends on the character encoding. For example, using ASCII encoding, the character "6" is represented by the binary number 00110110, which is the decimal number 54.
http://www.ascii-code.com/

So what you have is a text file that lists sample values, not "raw data".

In order to convert that text into audio, you would a program to "read" the text and generate sample values (the actual "raw data") from the text.

In this forum post there is some experimental Nyquist code for reading sample values from a text file: http://forum.audacityteam.org/viewtopic ... 13#p207413

Does that help?

[Sprinterdriver]

Thank you, I already know the theory behind different data types and text encdings, that is not the issue.

Your answer confirmed what I suspected - that I need to convert the text file by using another software - wich by the way doesn't seem to exist, at least due to web search for "convert text wav".

I'm a little above beginner level in Autohotkey, so I thing I would be able to save the values as a binary file, wich I plan to import using Audacity (after some research on their forum I hope to figure it out - that is another issue).

However - I hope someone here does know how to put together a raw audio file in order to be able to correctly import it in Audacity.

[cyrano]

Excel can convert numbers into binary.

There are even webpages who do that.

But I don't have a clue how to go from binary to audio. Up until now, I always supposed raw audio to be binary. If it's hex, any hex editor can convert from binary to hex.

Can't Sox help?

http://sox.sourceforge.net/

[steve]

Your answer confirmed what I suspected - that I need to convert the text file by using another software - wich by the way doesn't seem to exist

Please read all of my previous answer.

[Sprinterdriver]

Thanks Steve. I've spend this afternoon and the day before by testing the plugin you reffered to.

I was able to use convert a CSV file with multiple coloumns into several single coloumns CSV files, while adjust the number format using Libre Office Calc.
The big issues was to keep the numbers in output csv file from growing beyond ±1, but sorted it out (within Calc).

Second - there is no "Browse" button in the plugin gui, just a input field. For that I have a Autohotkey script that I've put in the sendto folder. It simply put the file path(s) to the clipboard (unlike copy file from within Windows explorer):

Code: Select all

If !%0%
ExitApp
Loop %0%
{
GivenPath := %A_Index%
i := A_index
Loop %GivenPath%, 1
{
If (i = 1)
LongPath := A_LoopFileLongPath
Else
LongPath := LongPath . "`n" . A_LoopFileLongPath
}
}
clipboard := LongPath
Sleep 100

Third - At my locale - the decimal separator is a comma (not a dot), so in aditional I need to do a search/replace for that in the CSV file.


Result / Bonus:
I've now the compressed samples of the physical G-forces my bike takes over a ten minutes ride, and I have used Audacity (among other programs) to convert those movements into auditible sound - as if the bike was a pickup.

The only not perfect thing about this is that the record isn't for the actual movement as for a vinyl pickup coil, but rather the acceleration. To fix this I tried to (in Calc) do a integration of the samples. At first samples (couple of pages) it looks good, but the integrated (converted to movement/distance) it slowly grows to very large numbers, and is therefore useless for converting to audio. Well - unless I find a way to simulate high pass filter in large data samples in Calc. The latter belongs to another forum - I just want to shed som light on what I've experienced given the help I got here.

[steve]

Second - there is no "Browse" button in the plugin gui

Unfortunately that is currently a limitation of Nyquist plug-ins. Hopefully there will be a browse button some time in the future, but for now we have to make do with an ugly workaround of using a text box. This limitation is the main reason that I've not pursued developing this plug-in further.

For that I have a Autohotkey script that I've put in the sendto folder. It simply put the file path(s) to the clipboard (unlike copy file from within Windows explorer):

Thanks for posting the Autohotkey script. I can't test it myself as I don't use Windows, but it may be useful to others.

Third - At my locale - the decimal separator is a comma (not a dot), so in aditional I need to do a search/replace for that in the CSV file.

As with most computer languages, Nyquist requires a dot as the decimal separator.
At the time that this plug-in was written it was not possible for Nyquist to detect the locale. Nyquist in Audacity can do that now, so it would now be possible to extend the plug-in code so that it automatically converted commas to dots where a comma is the decimal separator for the locale.

I've now the compressed samples of the physical G-forces my bike takes over a ten minutes ride, and I have used Audacity (among other programs) to convert those movements into auditible sound - as if the bike was a pickup.

Ha Cool
Please post a short sample, I'd love to hear how it came out (how to upload audio samples to the forum: http://forum.audacityteam.org/viewtopic ... 49&t=72887)

The only not perfect thing about this is that the record isn't for the actual movement as for a vinyl pickup coil, but rather the acceleration. To fix this I tried to (in Calc) do a integration of the samples.

An alternative approach would be to integrate the audio samples in Audacity.
Nyquist has an "integrate" function: http://www.cs.cmu.edu/~rbd/doc/nyquist/ ... l#index375
If you want some help with that, just ask.

Well - unless I find a way to simulate high pass filter in large data samples in Calc.

which of course is very easy to do in Audacity and/or Nyquist.

The latter belongs to another forum

If you'd like to discus more about using Nyquist, I can move this topic to the Nyquist part of the forum.

I just want to shed som light on what I've experienced given the help I got here.

Thanks, very interesting, and glad that you've had some success with it.

Excel can convert numbers into binary.

Converting into text notation of binary numbers does not help.

Up until now, I always supposed raw audio to be binary.

It is binary.

any hex editor can convert from binary to hex

No, a hex editor converts binary into text.

[Sprinterdriver]

Ok here is the sound:
https://soundcloud.com/random-stereo-so ... compressed
It was not ten minutes but rather five and a half - so it makes the "time compression rate" something like 189,5 times.

The fun thing is when I see the graphical wave form in Audacity, I recognise every spike as a speed bump or crossing of road and other known bumps on the route

If any interest I can try to make an ahk script to open a file within the nyquist gui box...

[steve]

Thanks Sprinterdriver, that's fantastic

[Sprinterdriver]

About the integration, I think there is a major issue - a large carrier wave bothe lenghtwise and values of max/min sample.

For a data serie where the samples lays between -1 an +1, and even if all samples is substracted to the average value (attempt to prevent growing) - I still end ut with a "long wave" that has a wave length that spans beyond the length of the data sample, but the largest values is like -300 and +480 - and I doubt that those data are accepded for sample input.

Therefore I've figured the best method is probably to filter this out somehow using Libre Office Calc. I've learn the basic about digital filtering ag school - almost twenty years ago - and cannot recall the methods by now.