Digital Signal Processing (DSP)

Once a signal has been decomposed into frequencies, the practical question is which helpers actually filter it, find its peaks, and detect beats. This page is a toolbox shelf: SciPy for time-domain signal processing and peak finding first, then beat detection with Librosa and a real-time detector, then two modeling notes on activity recognition and self-attention for sound.

The same notes are in Electronic Network Frequency Analysis and Feature Engineering.

The base layer is SciPy. Scipy signal processing is the signal processing module (scipy.signal) in the SciPy v1.18.0 manual, and Script find peaks is its find_peaks reference, the function to reach for when the question is where a signal spikes.

Peaks in audio lead directly to beat detection. Real time bpm beat det is a Python beat detector that does real-time detection of beats for audio, calculates BPM, and flashes an LED strip in time with music. Librosa also covers time-domain processing and beat detection (and temp), but its original docs addresses no longer open and are kept at the end of the page.

The same time-series thinking carries over to sensors. Mastery on Human activity recognition, smartphones is MachineLearningMastery.com on 1D convolutional neural network models for human activity recognition from smartphone data.

(Out of place) — Christopher Dossman's using self-attention for sound signal processing is Toward Interpretable Music Tagging with Self-Attention. Out place on a DSP shelf, it is a model rather than a signal helper.

Deprecated links

These links and images no longer work. The original wording is kept here. A same-resource copy, when one was checked, is used above.

• Librosa. This address no longer opens: https://librosa.org/doc/latest/core.html#time-domain-processing

• Beat detection (and temp). This address no longer opens: https://librosa.org/doc/latest/beat.html#beat