# How to differentiate an empty aerosol from a full one using only sound measurements?

The challenge I have is to detect an empty spray container using sound measurements only, since I can't place any sensor directly into the container because of the mechanism that tightens it. My main idea was to calculate the FFT of a 2-second recording just when the mechanism squeezes the trigger of the sprayer and then get the maximum peak frequency. However, I don't know if it is appropriate to use peak detection because when it is empty, it does not emit sound.

Any ideas for solving this ?

• Welcome to SE.SP, Jandry! Several of our users have flagged this post for closing. The reason they are giving is that your question is "off-topic". This reason is a bit too broad, but I think in the case of this question it's because you're asking two (mixed) questions: Can I detect whether an aerosol is full using sound measurements? (which isn't signal processing) and Will taking the FFT of a sound recording and finding the peak let me detect something? (which is signal processing). Please edit your question and focus on the 2nd question. – Peter K. Nov 27 '18 at 15:29

What I would do is :

1) Acquire the audio signal (or in fact, several, like ten) corresponding to a full container. I'd then quantify the energy of the signal, by squaring it and dividing by the number of samples. The obtained quantity is my reference energy. If several signals were measured, I take the average of calculated energies. Let's call it $$E_{full}$$

2) Do the same with an empty container (which is basically measuring ambient noise if it does not emit any sound like you say). Let's call it $$E_{empty}$$

From this point, you can define any energy threshold, with which you will compare the energy of the sound emitted by your experimental aerosol.

You can also do it simpler, by detecting for example the maximum peak amplitude of the full and empty aerosol and defining peak-amplitude threshold accordingly.

The threshold definitions can also be fine-tuned by measuring other aerosols with varying fillings, like half-empty, 25 % etc...

And finally, all the measurements I suggest implies that the sound is measured in steady state.