The effect of Gaussian fitting is not good

Question

I am working on a gaussian fitting problem. The data distribution is like follows:

There are four distributions, so I use four different params for each distribution