You may use a CA-CFAR which is a technique initially used in radar systems that has been applied in many other fields. It goes like this: you decide if a sample is an outlier by comparing with the average of the nearby samples. You must multiply this average by a given factor before executing the comparison. Choosing a high factor value will result in a low the false alarm probability. On the contrary, a low factor will provide a high probability of detection. I suggest you to start with a factor of 3 and to use the 60 samples around the one being analyzed for the average calculation.