Make each voice track a different frequency for band filter experiment

Greetings,

Apologies, I am not a musician and know nothing of the phrases ‘pitch/tone/etc.’ I am a beginner EE student learning about filters.

What I want to do is: build several bandpass filters with a very narrow allowable band, say 5kHz. I will then record several different tracks of my voice saying different sayings. One track will be 5kHz; every other track will play at some other odd frequency.

I want it to sound like noise if played unfiltered; then pass the signal to my filter and have it attenuate all other frequencies, thereby playing my “secret phrase” coming at the right frequency.

Is this possible? (Is this not how the original Bell Telephone worked?) There’s gotta be a way, but I don’t know how to proceed.

Currently, the best way to get a steep band filter, is to use the “Filter Curve” effect (https://manual.audacityteam.org/man/filter_curve.html)
Drag the lower slider on the right hand side down to the bottom so that the dB scale goes down to -120 dB.

This image shows a band-stop filter for a frequency range of about 400 Hz to 1000 Hz.
Experiment away.

filter-curve.png

A couple of problems -

Your voice contains a wide range of frequencies. Analyze → Plot Spectrum will show you the frequency spectrum. If you narrow the spectrum too much it simply won’t be intelligible.

And, 5kHz isn’t “narrow band”. i.e. The bandwidth of the regular landline telephone system is something like 300Hz to 5kHz.

I want it to sound like noise if played unfiltered; then pass the signal to my filter and have it attenuate all other frequencies, thereby playing my “secret phrase” coming at the right frequency.

This is a different experiment, but you can try mixing your voice at a low-level with white noise, pink noise, or music so that your voice is masked (drowned out). Keep a copy of the original noise/music file and invert it. If you mix it back in the two noise files will cancel leaving the voice. And, you can re-amplify the voice if it’s too quiet.

I don’t think masking one voice with other simultaneous voices is going to work …

Digital signals offer greater opportunities to hide data securely,
e.g. MP3Stego.

Thank you for the ideas this far, I really appreciate it.

I failed to clarify earlier: I said I wanted to use a band of 5khz. What I meant to say is that I would make a track (a secret phrase or something) that only outputs frequencies close to 5kHz, plus or minus however many. Several other tracks would be recorded the same way, but centered around other frequencies.

My filter would be a very narrow bandpass around that 5 kHz. I suppose I want to try to make the band as narrow as possible.

Is this a bad idea? I thought about doing something similar with a midi piano and Morse code… Think that would be better?

Your best option currently is to use the “Filter Curve” effect.

The next Audacity release has a new filter that will allow even narrower bands than are available from the Filter Curve effect, but that will not be available for about a month.

You need at least 4kHz bandwidth for intelligible speech.

Even if you interleaved the spectra of two voices,
I think it would be possible to understand both without filtering-out every-other (narrow) band.