Locomotion Mode Recognition Using Sensory Data With Noisy Labels: A Deep Learning Approach

Availability of various sensors in the smartphone makes it easier and convenient to collect the data of human locomotion activities. A recognition approach can utilize this sensory data for recognizing a locomotion mode of a user such as a bicycle, bike, car, etc. Such recognition of locomotion modes helps in the precise estimation of transportation expenditure, travel time, and appropriate journey planning. The accuracy of the recognition approaches heavily relies on the training dataset having correctly annotated labels. These labels are usually assigned using crowdsourcing or web-based queries for economic and fast annotation. However, the annotation generates abundant noisy labels in the dataset. This paper proposes a locomotion mode recognition approach capable of handling noisy labels in the training dataset. The approach builds an ensemble model by developing three different deep learning-based models, namely conventional, noise adaptive, and noise corrective, to handle different concentrations of noisy labels. The ensemble model not only improves the recognition performance but also helps in estimating the concentration of noisy labels. Experimental results demonstrate the effectiveness of the proposed approach on collected and existing datasets.

Automated summary

The availability of various sensors in smartphones makes it easier and more convenient to collect data on human locomotion activities. A recognition approach can utilize this data to recognize the user's mode of locomotion, such as cycling, biking, or driving a car. This mode recognition helps in accurately estimating transportation expenditure, travel time, and planning journeys. However, the accuracy of recognition approaches depends heavily on correctly annotated labels in the training dataset, which can be noisy due to annotation methods. This paper proposes a locomotion mode recognition approach that can handle noisy labels by building an ensemble model using three different deep learning-based models.