CoDEm: Conditional Domain Embeddings for Scalable Human Activity Recognition