How to avoid detecting small and blurry face with haarcascade?

edit

Hi

I'm making a Facial Recognition program that uses IP Cam to recognized people in my office. I put the camera near the entrance door. The program uses opencv and haarcascade to detect face and once face is detected, it will send the face to AWS Rekognition to be recognized. However, the program keeps on capturing blurry and small faces. The "small" and "blurry" image is due to people entering the camera frame and moving towards the camera. By the time the recognition process finish, the person already left the camera view.

So my question is, is there any way for OpenCV or Haarcascade to set a minimum width of face that it detects?