Optimizing filter to give certain local minima points

I have a plot of average speed versus time with some noise. The average speed drops to about zero(stationary) at certain times. How can I can optimize my filter/ or smooth function to only 9 minima points, as there should be only 9 stationary points in original data. Thanks

[Image Analyst]

Do you mean that if your original list (vector) of speeds has 10 or more zeros in it, you want to remove those elements from the vector so that only the first 9 zeros (plus non-zero data of course) remain?