Matt,
I've spent a lot of time with SPASM; it's impressive and I think it has a
lot of potential, but I've not yet been able to achieve the
performance/quality I need in analysis. Perhaps, that's on my end with some
lack of understanding with the tool and/or improper configuration values.
My home grown tracking software works better and I don't think that should
be the case.
Thanks for working on a 2.0 version and here is my wish list!
Documentation with detailed explanation of SPASM's framework, components,
models, methods, parameters, and their relationships to one and other.
Plenty of usable examples solving a variety of real-world problems,
including adaptive-hybrid and composite filters. For example, I'm
interested in realtime linear and non-linear path tracking, prediction and
the smoothing of stochastic signals that contain gaussian addative noise
(but not plane trajectory modeling or GPS synchronization).
Explanation of why and how to configure and interpret the parameters of the
state transition model, process noise model, observation model, and the rest
(with configuration, initialization, and manipulation helper functions).
Velocity and acceleration are used quite a bit; it would be nice to have
functions/objects, which encapsulate these and other values of importance on
and across multiple scales concurrently.
Likelihood is very powerful and should be improved, evolved and explained.
Filterbank is also a great concept and powerful feature that can evolve,
should be documented, and could be a showcase for advanced-complex filtering
and the software's analysis framework.
Operational performance is slow in critical areas and should be tweaked,
tuned and enhanced.
Best Regards,
Norman Bates
|
