fCWT application question about fcwt

[fastlib]

Hi, thans for your questions!

• For each dataset and sampling rate, we calculated the Realtime Analysis Ratio (RAR) which is the ratio between the signal's time span and duration of computation. With a RAR < 1 the computation is faster than the signal's duration which means that real-time analysis is within reach. Currently, fCWT is not yet programmed to work with datastreams. However, fCWT can be easily adapted to that application if a signal is split up into chunks that are just wide enough to encapsulate the lowest frequency one wants to measure (i.e., a 10Hz signal requires a window size of 0.1s and a 0.5 Hz signal needs a window of 2s). Each chunk is analysed by fCWT. As this analysis is much faster than the chunk's duration, one obtains real-time performance. Fortunately, this does not decrease time-resolution as you say. Within each chunk, the time-resolution is still at the per-sample level as would be in offline research. You could even dynamically adapt the chunk length to each scale to reduce unnecessary latency in the high-resolution spectrum.
• The application of fCWT to real-world on-line applications would be a great topic for future research!

I hope I have answered your questions.

[sunshineinsandiego]

Yes, perfectly! Thank you

[stellarpower]

So, to confirm - the implementation is only realtime/online in the sense that it is fast enough to be called repeatedly? It doesn't keep some sort of internal state and recalculate when I e.g. push it a sample or small packet of samples, I would need to implement a sliding window myself, and call the function repeatedly over that window?