Question: Is it possible to generate purely synthetic fart sounds?
Short answer: Yes.
In the past, we have devised a way of creating infinite varieties of new farts out of combinations of old farts. While these are somewhat interesting, they hardly constitute purely synthetic farts, being comprised of mixtures of pre-existing biological samples. However, even in the absence of any such data, it is possible to produce mathematical descriptions of the acoustic properties of the human rectum and anal sphincter. We hypothesized that combining these elements could lead to an entirely synthetic approximation of fart sounds.
Building on the popular source-filter model of speech production, we characterized the input to the fart production model as broadband noise. We then filtered it according to the acoustic properties of the rectum and sphincter. These have specific resonances that lead to sounds with a peak frequency of around 250 Hz, as well as smaller peaks at the odd harmonics. Here is an example of noise filtered at these frequencies:
Clearly, something is missing.
It turns out that the acoustic properties of farts, like speech, cannot be captured simply by the constituent sound frequencies. Rather, they are strongly influenced by something called Bernoulli’s Principle, named after the eighteenth-century mathematician Daniel Bernoulli.
Briefly, the Principle states that static air pressure decreases with air velocity. So when air rushes out of the anus during a fart, the pressure outside the anus decreases, causing an influx of air that pushes the anal sphincter shut. Pressure then builds up from the rectum, opening the sphincter again. This rapid opening and closing of the sphincter causes fart sounds to be released in pulses; precisely the same phenomenon occurs with the vocal folds during the production of speech.
Bernoulli’s Principle is apparent in the pulsatile nature of real fart waveforms, as can be seen in this example:
The corresponding audio is here:
To incorporate the Bernoulli effect into our fart production model, we simulated a series of random pulses, which we then used to modulate the filtered noise input described above. The full model is shown here:
Here is an example of a purely synthetic fart generated in this way:
We consider this effort to be a limited success, as it reproduces the general sound characteristics of farts, as well as their pulsatile waveform:
Our fart detection algorithm, FartNet, gives this sound a 61% chance of being a fart, which is not bad for a first attempt. But like other, less interesting attempts at synthetic biology, there is still much room for improvement. To be continued.