How to easily extract spike times related to a recording chunk.

[Djoels]

Just a question: how can one easily create selection of a time fragment along a recording.
For the recording you can make a call to select_channels and time_slice, to get a chunk of the recording.
However when trying to extract the affiliated spike times to this time_slice, I find I'm having to play with the sampling frequency (given that I don't have a time vector).

Am I missing some elegant way of doing this?

[zm711]

I always just do this myself by converting to samples from time (as you said using the sampling_frequency), but maybe @alejoe91 or @samuelgarcia know of a machinery in spikeinterface to easily do this (in which case I might switch myself :) )

[chrishalcrow]

Hello! There's a few ways to look at spike trains: to_spike_vector, get_unit_spike_train and I recently found spike_vector_to_spike_train(!). Of these, the get_unit_spike_train method has a return_times parameter. So if you have a sorting object called sort you can do something like

unit_id = 5
unit_spike_train = sort.get_unit_spike_train(unit_id = unit_id, return_times=True)

t_start = 1
t_end = 4
interval_train = unit_spike_train[(1 < unit_spike_train) & (unit_spike_train < 4)]

But if you're working with the to_spike_vector output, you'll need to deal with the sampling_frequency yourself.

[alejoe91]

Another option is to use the sorting.frame_slice() to get a new sorting object. I think we could add a time_slice, which automatically does the conversion/checks the registered recording for timestamps.

@chrishalcrow can you give it a try?