and walking downstairs, respectivel

and walking downstairs, respectively. We only show the resultsforPWAscenario in Table IV and not forPWAGbecause theyhave almost same recognition performance. The performanceimprovement of adding the two additional gyroscopes is notsignificant, especially taking into account its impact on thebattery life, which is yet to be seen in practical implemen-tations. Using only smartwatch sensors performs worse thanits combination with the smartphone sensors. Therefore, it isbetter to combine these two devices for better recognitionof these complex activities. Within smartwatch sensors, thecombination of gyroscope and accelerometer performs betterthan their individual performances for some of the complexactivities, such as eating, drinking coffee, giving a talk, walk-ing upstairs and walking downstairs. The performance resultsof these different combinations using SVM, IB1 and decisiontree classifiers are shown in Table IV. We do not show theresults for sitting, standing, jogging, and biking, because theirresults are very high and we observe negligible performanceimprovements for them in different fusion scenarios. In gen-eral, the recognition performance of using the gyroscope alonewas poor and that is why we do include it in these graphs.Moreover, the recognition performance of complex activitiesat pocket position alone was poor, so we did not report theseresults