Operations
Each page covers a family of operations: what they do, the SciPy call they match where there is one, and how to run them from Python, Rust, a pipeline spec or the app.
| Family | Operations |
|---|---|
| Filters | IIR, notch, SOS, FIR, Gaussian, median, Teager–Kaiser, template subtraction |
| Referencing | CAR, CMR, Laplacian, whitening |
| Rate | resample_poly, decimate |
| Per-sample math | scale, baseline, clamp, rectify, square, sqrt, log10 |
| Channels | Channel maps, reordering, selecting |
| Statistics and models | Noise levels, windows, peak candidates, PCA |
| Pipelines | Running chains over whole recordings |
Everything runs on the device
Section titled “Everything runs on the device”Data is (channels, samples). Operations run on the runtime
you pick, and return float32 of the same shape, unless the operation changes it.
Conventions
Section titled “Conventions”- Sample rate. Design functions take it last; functions applied to data take it after the data. Stages take none, because the pipeline knows the sample rate.
- Defaults are shown in every signature, and follow SciPy where an equivalent exists.
- Rust examples below are stages for a
Pipeline. Each family also hasexecute_*functions on device buffers.