krang | The Windows desktop client for Hadouken v5 | Dektop Application library

Handle the result of the MsgBox:

This is an Audio Event Detection problem. If the sound is always the same and there are no other sounds at the same time, it can probably be solved with a Template Matching approach. At least if there is no other sounds with other meanings that sound similar.

The simplest kind of template matching is to compute the cross-correlation between your input signal and the template.

1. Cut out an example of the sound to detect (using Audacity). Take as much as possible, but avoid the start and end. Store this as .wav file
2. Load the .wav template using librosa.load()
3. Chop up the input file into a series of overlapping frames. Length should be same as your template. Can be done with librosa.util.frame
4. Iterate over the frames, and compute cross-correlation between frame and template using numpy.correlate.
5. High values of cross-correlation indicate a good match. A threshold can be applied in order to decide what is an event or not. And the frame number can be used to calculate the time of the event.

You should probably prepare some shorter test files which have both some examples of the sound to detect as well as other typical sounds.

If the volume of the recordings is inconsistent you'll want to normalize that before running detection.

If cross-correlation in the time-domain does not work, you can compute the melspectrogram or MFCC features and cross-correlate that. If this does not yield OK results either, a machine learning model can be trained using supervised learning, but this requires labeling a bunch of data as event/not-event.

The length of time is inversely proportionate to the magnitude of dt. And you have a nested loop based on the length of time in each case. As long as that loop is your hottest code, your code will experience quadratic growth (O(n**2)). Stepping from 0.01 to 10 by 0.01 means a length of (close enough) 100 elements, and ~10,000 units of inner loop work. Doing the same thing from 0.001 to 10 by 0.001 means ~1000 elements, and ~1,000,000 units of work, an increase of 100x.

From a starting point of 38 seconds, that's pretty extreme; even if memory weren't an issue, you'd be looking at 3800 seconds (over an hour). And memory is an issue; your inner loop repeatedly appends to diff_i, so you're going from storing ~10,000 floats (which on CPython x64 builds seem to occupy 24 bytes a piece, plus 8 bytes for the reference in the list), which means you eat about a third of a MB of RAM. With dt of 0.001, that ends up closer to 32 MB, and going to 0.0001 would bring you up to 3.2 GB. Even if you have that much RAM, it means you're no longer benefiting from CPU caches for a lot of stuff, which will likely slow you even more than the straight CPU costs would indicate.

Your code takes very little advantage of numpy's features, and could likely be tightened up quite a bit. Doing so would save a great deal of memory, and allow most of the work to be pushed to single function calls with the loops performed in C, running much faster than the Python interpreter.

For the simplest improvement, take diff_i. Before the loop even runs, you know exactly how many elements it will have, and the calculation could be simplified to simple array operations, converting j_i to a numpy array before the outer loop even begins, and replacing the loop with a simple series of computations on j_i that directly produce diff_i as a numpy array with no Python level loops at all.

I haven't tested this (no numpy on this machine), but a rough stab at this would be to convert j_i immediately after it's been populated with:

The answer was staring me in the face.

It was unwise to split the adjust Y functions into two separate functions, since they both relied on one another to be completed simultaneously. I still don't know precisely why Javascript decided to stop after 3 iterations, but my guess is that one function "got ahead" of the other around that time and caused a discrepancy, so one function was multiplying a non-existent value, resulting in NaN.

Here is the updated code for anyone in the same boat as me:

Try this. I've added a worksheet variable ws. This puts the sheet name in column A of the new workbook, and the data in col B onwards. I have also added declarations for all the variables.