Intuitively, human memory compares the past with the present to decide if something has changed in the environment. It is additionally finite and causal. The idea of memory is central to both our audio and video scene segmentation algorithms. We implement memory a finite causal buffer with two parameters:

1. Memory: The total data in the buffer (typically around 30 sec.)
2. Attention span: The most recent data in the buffer. This data is used to compare against the rest of the memory in order to decide if whether a scene change has occurred.