Function after spline fitting

Question

Balint Egri 2021-3-26

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/784311-function-after-spline-fitting

评论： Balint Egri 2021-4-1

Hello hello,

I have a question regarding spline fitting. I have the below curve which obtained by fitting a curve on my test data. (this is the roll angle of an IMU data) I used spline fitting and piecewise polynomial to have it smoothend by increasing the increment (time) 10 times.

I want to use this data in another program to create motion on a body.

Is it possible to obtain this curve az a sin/cos function? Or any kind of function?

Thank you for your answers:)

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Balint Egri 2021-4-1

I attached here the .mat file where I am doing the curve fitting. I also attached the .txt files which the spli fitting reads. Hope this can help.

Do you think the fft could work with this set of data?

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

John D'Errico 2021-3-26

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/784311-function-after-spline-fitting#answer_659111

You cannot extract any "function" from a spline. At least, if you did, it would not be anything you could read or type without many, many coefficients, that you problably would not find useful.

In order to build an approximation using a sum of sin and cosine terms, you will usually use a tool like the curve fitting toolbox, but the optimization or stats toolbox can also do nicely. you can also do it without any such toolbox, but it wil take more effort. So knowing what you have would be important. Also, it would greatly help to have your data.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Balint Egri 2021-4-1

I commented on the post with my data. Check it out if you think the curve fitting is a working solution:)

请先登录，再进行评论。

Answer 2

Star Strider 2021-3-26

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/784311-function-after-spline-fitting#answer_659156

The easiest way to fit it to a sin or cos function (or both) to it would be to take the Fourier transform of it using the fft function. It would then be possible to reconstruct it in the time domain from the real and imaginary coefficients at the frequencies you want to use for that. (It would likely require several terms to get a decent approximation to the data.)