Faster sliding window statistics?

Question

K E 2016-10-13

1
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/307202-faster-sliding-window-statistics

编辑： K E 2016-10-14

I use slidefun to estimate statistics such as max, min, or RMS within a sliding window applied to a time series. It is very useful, but it can be slow if there are a lot of data points. Is there a faster sliding window routine out there? I didn't see any obvious candidates in the File Exchange, but would like to know if I missed any.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Steven Lord 2016-10-13

3
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/307202-faster-sliding-window-statistics#answer_238882

If you're using release R2016a or later, consider using the moving statistics functions in MATLAB for at least min and max.

Depending on exactly what type of windows you're using, if you're using release R2016b storing your data in a timetable and using the retime function with an aggregation method is another potential option.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

K E 2016-10-14

编辑：K E 2016-10-14

On my machine: with movmin , 0.7s. With slidefun, 310s. This is really going to streamline a lot of work for me & my coworkers.

请先登录，再进行评论。

Answer 2

Image Analyst 2016-10-13

2
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/307202-faster-sliding-window-statistics#answer_238887

You can use these alternate functions. For mean, use conv(). For max, use imdilate(). For min use imerode(). Or the new moving stats functions Steve mentioned. They're all highly optimized. Whether they're faster than slidefun() is something you'll just have to check.