I have purchased a set of MP3 files (with classical and binaural music) that claims to have embedded in them messages at a frequency that people cannot hear -- i.e., subliminal messages (these are for example motivational messages). How do I know whether the seller/author is telling the truth and indeed that these MP3 files do contain these positive messages at frequencies that humans cannot hear? How can I reverse-engineer to hear only the spoken messages, stripping out the audible music?
Hiding messages in an unhearable spectrum will have zero effects on any human, no matter what they claim. However, we can still analyse whether they have indeed hidden something or not.
You could use the free tool Audacity to do all this.
The MP3 encoder usually heavily compresses anything above 16 kHz or even completely remove it, depending on the bitrate. Whatever you will hear may be hard to understand, unless the bitrate is high.
Check the sample rate of the files and divide it by 2. That is the maximum frequency that can be stored in the file. The sampling rate is probably either 44.1 or 48 kHz, higher rates are not supported by MP3. This gives you 22.05 or 24 as the maximum frequency.
First, apply a high-pass filter to get rid of the parts you don't want to hear. Adjust the value such that it is between the music and the subliminal messages. It might be a bit hard to find that point, especially if the hidden messages frequency band is quite near the music frequency band, then it might be a good idea to set it to 12 kHz. That way you will get rid of pretty much all of the music and not risk to accidentally removing something of the hidden parts.
Now, apply a pitch shift effect to transform the maximum frequency down to the audible spectrum. This depends on your age. Young people usually can not hear above 18-20 kHz, older people often can barely hear 16 kHz. Of course you can shift further down, like down to 12 kHz. Beware, this means the maximum frequency will be at 12 kHz while everything else is below that value, shifted accordingly by the same amount. This means, if you shift further down, at some point, the lower-frequency parts of hidden messages may be pushed into a spectrum that is too low to hear.
And then you're done. Listen to it. If the voice is too low you could undo the first shift and redo it with a somewhat smaller value. But if you hear a high pitched voice, do another round of pitch shifting, this time with a much smaller value.
that claims to have embedded in them messages at a frequency that people cannot hear
That's an odd claim to make. Subliminal messages can be done in video: the theory is that you can show a message during 1 frame of the video and it'll be too short to be consciously picked up, but it can be read and processed unconsciously.
If an audio message is at a frequency humans can't hear, they can't process it at all, subconscious or otherwise. It's more likely the message is at a level where it's masked by the music: when the level is only a few dB lower than that of the music, humans already can't hear it.
But as I understand it, you can't do that in MP3: it compresses audio by removing parts that we can't hear, including sounds that are at a lower level than the dominant sound. So the conversion to MP3 would strip out those subliminal messages.
You can't strip out the music entirely, but you could filter out the lower frequencies (below 200 Hz, above 4 kHz), this leaves the frequency range of human speech intact and might enable you to hear the messages if they're there.