Human activity recognition (HAR) is an emerging methodology essential for smart homes with practical applications such as personal lifecare and healthcare services for the elderly and disabled people. In this work, we present a novel HAR methodology utilizing the recognized body parts of human depth silhouettes and Hidden Markov Models (HMMs). We first create a database of synthetic depth silhouettes and their corresponding body parts labelled silhouettes of various human activities to train random forests (RFs). With the trained RFs, a set of 23 body parts are recognized from incoming depth silhouettes, yielding a set of centroids from the identified body parts. From the dynamics of these centroids, motion parameters are computed: a set of magnitude and directional angle features. Finally, the spatio-temporal dynamics of these motion features of various activities are used to train HMMs. We have performed HAR with the trained HMMs for six typical home activities and obtained the mean recognition rate of 97.16%. The presented HAR methodology should be useful for residents monitoring services at smart homes.