hpr-knowledge-base/hpr_transcripts/hpr1343.txt

Episode: 1343
Title: HPR1343: Too Clever For Your Own Good
Source: https://hub.hackerpublicradio.org/ccdn.php?filename=/eps/hpr1343/hpr1343.mp3
Transcribed: 2025-10-17 23:52:58

---

.
Hello, this is Leandere for Hacker Public Radio, recording 20th August 2013.
Two clever for your own good.
Today I'd like to talk to you about an experience I had that involved my love of programming
and Hacker Public Radio.
Let me set the stage.
The date is April 1, 2013, and Ken Fallon has just released Hacker Public Radio
episode 12-16 digital data transfer.
As some of you may recall, this episode was nothing more than several minutes of Morse
Code.
Now there was a time when I spent a great deal of time getting better at translating
Morse Code, but I haven't stayed in shape and it would have been a lot of work for me
to sit through this episode and translate it by hand.
So what I decided to do instead was find a way for the computer to translate it for me.
Many of you are probably aware that Linux comes with a program that handles translation
to and from Morse Code, named Morse, but it only works with converting ASCII text of
DIT and DAW respectively into an out of English text.
So how was I going to translate an actual audio file into the corresponding DITs and DAWs?
I thought of a few different possible ways to do this, looking at programs that would convert
the audio into a waveform, in some kind of an image file, and then parsing the image
file for periods of sound and silence.
I looked around for a few different packages and really couldn't find anything that would
do the work I needed it to do, and that's when I suddenly remembered that really uncompressed
audio is just a numeric representation of the waveform.
So if I could just convert it into uncompressed audio, I would probably be able to output it
in some kind of a format that I could use.
So I started by converting the org file into an uncompressed wave file using SOX.
SOX is billed as the kind of audio Swiss Army knife.
So I just started with SOX, hpr1216.org, hpr1216.wave, which would just use the defaults
and convert the file.
After having done that, I needed to know some of the information about the internal structure
of the wave.
So I used SOX, another tool that comes with the SOX suite, to get the metadata about
the wave.
So SOX, hpr1216.wave, and that gave me a lot of information that I needed.
The audio was a single channel, that is mono.
The single rate was 44,100 samples per second, encoded as 16-bit signed integers.
So really, this gave me all the information I needed, and now I just needed to access
the actual samples and spit them out in a way where I could detect sound and silence,
preferably in a line-by-line format, so I could feed it into whatever text processing
script or whatever I could manage to come up with.
Now I could have gone probably straight to raw audio, but since I went to wave, because
I was familiar with it, I think figured it would probably be useful to remove the wave
header.
So first I needed to know how long that was, an easy way to do that is just to generate
an empty wave file.
So I broke out SOX again and went with SOX-T raw for the input type, dash B16 for 16-bit
samples, dash R44100 for 44,100 Hertz sample rate, dash C1 for single channel audio,
mono, and dash E signed hyphen integer to get the appropriate encoding, slash dev slash
null as my source file of type raw, and empty dot wave as my output file.
Now running that command just spits out a single 44-bit file, which is just the wave
header.
So now I had the information that I needed to be able to skip over the wave header in
hpr1216.wave.
Next step is to get it into some kind of usable text format.
Now you're thinking of converting binary data into text, the first few things that come
to mind are maybe base64 or hexadecimal encoding.
I opted for the latter, since I wouldn't have to worry about variable length, numbers
would all be very easy to work with.
So I broke out one of my favorite tools, hex dump, and the command line I used for this
was hex dump, hyphen S44 to skip the first 44 bytes, hyphen V for verbose that is not
to drop repeated sequences, for example long runs of zeros, and then format strings, hyphen
E single code 220 slash 2.
So that is 220 repetitions of 2 bytes each, double quote %04 lowercase x double quote.
So that is going to take an output for hex digits for the sample that's being encoded.
So again, that's taking a 16-bit signed integer and outputting it as 4 hex digits.
And a single quote to close that format string, hyphen E for an additional format string,
single quote, double quote, back slash lowercase n, double quote, single quote.
And so that's going to output a line ending after each run of those 220 samples, hpr1216.wave,
greater than sign, hpr1216.hex.
So what that's going to do is take 220 samples, output it on a single line, repeat this for
the whole file, and spit it out into my output file.
Now I tried this with an even 441 samples, which would have been 10 milliseconds of audio
per line, but I had too much overlap at such a high resolution between sound and silence,
and the eventual result I got was just too garble to work with.
So the next thing I'm going to do is to find all the currencies of silence and replace
them with something that's a little easier to detect visual.
For this I used said, and the command said hyphen E, single quote, s slash 0, 0, 0 dot slash
space, space, space, space slash g, single quote, hyphen E, single quote, s slash ff dot
slash space, space, space, space slash g, single quote.
So what that's going to do is it's going to take all the currents of either 0, 0, 0 followed
by a single character, and fff followed by a single character, and globally replace those
with four spaces.
Since we're working with signed integers, that means not only is values near 0 close to
silence, but also values near negative 1, which is ffff.
So taking my hex file, I use that said command, hpr1216.hex, greater than sign hpr1216.space.
So now this gives me a file full of essentially random hex digits separated by long runs of
lines filled only with the space character.
So if you open that with the less command and use hyphen capital S to not wrap lines,
you can actually scroll through the file and visually see the dots and dashes as short
runs of non-white space lines, followed by long runs of non-white space lines.
This I thought would be a good enough format that I could run it through a text processing
tool and actually get dits and does respectively.
So my favorite programming language for this kind of operation, and really just any quick
one off prototype program, is AUK.
And since this is really text processing, this is really where AUK shines.
So my AUK program basically had to do this.
It had to find lines that were full of just spaces, interpreting those as silence, and
then keep track of how many of those occurred in a row and how many non-white space lines
in a row.
So essentially I structured the program as a simple state machine.
It has four rules.
The first is basically to check to see whether the current line is all white space.
So the regular expression I used for that is slash, carrot, space, star, dollar sign slash.
So that's just saying find me runs of zero or more spaces that stretch all the way from
the beginning of the current line to the end of the current line.
I match that against the current line and store that result in a variable.
So essentially just a one or a zero of whether or not the line was silenced.
This first row also has to store the state of the previous line since we need to be able
to detect transitions between silence and non-silence.
So that rule looks like last equals this.
This equals dollar sign zero for the variable holding the entire line, tilde for regular
expression max, match, and the regular expression I previously mentioned.
The next rule checks to see where last is equal to this, i.e. we haven't changed state.
And if that's the case, increments the duration variable.
So this just allows us to keep track of how long we've been in the current state.
The last rule checks for not last that is the last line was not silence and this, i.e.
this line is silence, so transitions from sound to silence, which is going to tell us that
we just finished a run of some amount of sound.
So in that rule, I check if duration greater than 10 and duration less than 20 print
F period, which is the sign that Morse uses for a DIT.
Else if duration greater than 30 and duration less than 40 print F hyphen or the DAH character.
So whether those got matched or not, reset the duration to zero since we just changed states.
The last rule checks for the opposite state transition.
So going from silence to sound and the match criteria for that is last, i.e. the last
line was silence and not this, i.e. this current line is not silence.
This rule looks very similar to the last except the duration ranges are slightly different.
Here I check if duration greater than 30 and duration less than 40 print F backslash
N. That is to output a single line break to mark the end of a letter.
Else if duration greater than 80 print F backslash N backslash N.
So two blank, i.e. two line endings or one blank line to act as a word separator.
And again reset the duration to zero.
Next you take and run your dot space output, which again is those lines of hex characters
and lines of white space through this awk script using awk hyphen F, hpr.awk, hpr1216.space,
greater than hpr1216.dot.
So that gives you a file that contains a long series of periods and hyphens separated
by blank lines representing the actual Morse code in ASCII text.
This is suitable to run through the Morse program with the hyphen D or decode flag and
actually output something resembling human readable English text.
Obviously there are a few typos but really this was far better than I expected this to work.
All that was needed was to run it through a spell to correct some of the typos and the
text was easily recognizable as a portion of the Wikipedia page for Morse code.
So there you have it by being a little clever than you probably ought to be and doing things
that are a complete waste of time.
You can be lazy and teacher or computer to do something that you don't feel like taking
the time to do.
Thank you and see you next episode.
You have been listening to Hacker Public Radio at Hacker Public Radio.
We are a community podcast network that releases shows every weekday Monday through Friday.
Today's show, like all our shows, was contributed by a hpr listener like yourself.
If you ever consider recording a podcast, then visit our website to find out how easy
it really is.
Hacker Public Radio was founded by the digital dog pound and the economic and the computer
club.
We are funded by the binary revolution at binref.com, all binref projects are proudly sponsored
by LUNA pages.
From shared hosting to custom private clouds, go to LUNA pages.com for all your hosting
needs.
Unless otherwise stasis, today's show is released under a creative commons, attribution, share
a life, lead us our lives.