I have a time series analysis algorithm that uses arima modeling and fits a model to the time series and gets the residual values. It is more of a statistical signal processing algorithm than a digital one. I have a matlab code for it which uses functions from econometric toolbox.
I need to program it on an embedded system. My first instinct was to do this on dsp embedded system, but I couldn't find anything about parametric methods in the dsp embedded books. Most of the books only cover the FIR, IIR and adaptive filtering. Is it possible to do something like this on embedded systems?
and if yes, assuming that I use one of ARM cortex M series microControllers, is there any specific header file that I can use?
I hope my question is clear enough but I can explain it more.
I was trying to say that what I have is more into parametric estimation part of digital signal processing but I guess I said it the wrong way.
I have a machine learning algorithm (more specifically an outlier detection algorithm). my input is a time series data, lets say from an ADC and it is continuous but I am doing segmentation on the data. I think the segmentation will also help with the computational complexity and the limited RAM resources (If that was what you meant by estimating required RAM resources).
The limited resources of microcontrollers is my exact problem. I do have libraries in both C++ and python that will provide me with functions that I need for my code, but adding them all in the code will make it so huge. and that is why I thought maybe dsp systems are my best choice.