You could pre-compute the mean and std of your data sequentially using this approach. Would that work for you?