SpatialFilter#

class myoverse.transforms.SpatialFilter(kernel='NDD', grids='all', dim='channel', **kwargs)[source]#

Apply spatial filtering using grid layouts.

Spatial filters use 2D convolution on electrode grids. Grid layouts must be stored as a tensor attribute (via myoverse.emg_tensor).

Parameters:
  • kernel (str | torch.Tensor) – Filter kernel. Either a name (“NDD”, “LSD”, “TSD”, “IB2”) or a custom 2D tensor.

  • grids (str | list[int]) – Which grids to filter. “all” for all grids, or list of indices.

  • dim (str) – Channel dimension name.

Examples

>>> import myoverse
>>> emg = myoverse.emg_tensor(data, grid_layouts=[grid1, grid2])
>>> ndd = SpatialFilter("NDD", grids="all")
>>> filtered = ndd(emg)

Methods

__init__([kernel, grids, dim])

_apply(x)

Apply the transform.