Abstract—Recently, there has been a growing interest in theresearch community about using wrist-worn devices, such assmartwatches for human activity recognition, since these devicesare equipped with various sensors such as an accelerometer anda gyroscope. Similarly, smartphones are already being used foractivity recognition. In this paper, we study the fusion of a wrist-worn device (smartwatch) and a smartphone for human activityrecognition. We evaluate these two devices for their strengthsand weaknesses in recognizing various daily physical activities.We use three classifiers to recognize 13 different activities, suchas smoking, eating, typing, writing, drinking coffee, giving a talk,walking, jogging, biking, walking upstairs, walking downstairs,sitting, and standing. Some complex activities, such as smok-ing, eating, drinking coffee, giving a talk, writing, and typingcannot be recognized with a smartphone in the pocket positionalone. We show that the combination of a smartwatch and asmartphone recognizes such activities with a reasonable accuracy.The recognition of such complex activities can enable well-beingapplications for detecting bad habits, such as smoking, missinga meal, and drinking too much coffee. We also show how tofuse information from these devices in an energy-efficient way byusing low sampling rates. We make our dataset publicly availablein order to make our work reproducible.