Precise detection of the current activity status for chronic diseases patients could
play a significant role for protect their lives against sudden decline in health.
Combining the information form various data resources present a reasonable challenge.On
the other hand, stream classification techniques have a privilege of low computational
time but they need a feedback for adapting the classifier. This work proposes a model
for providing efficient automatic feedback for adaptive random forest classifier using
deep learning classifying of video stream from surveillance systems.