Accurate activity recognition in a home setting

Abstract

A sensor system capable of automatically recognizing activities would allow many potential ubiquitous applications. In this paper, we present an easy to install sensor network and an accurate but inexpensive annotation method. A recorded dataset consisting of 28 days of sensor data and its annotation is described and made available to the community. Through a number of experiments we show how the hidden Markov model and conditional random fields perform in recognizing activities. We achieve a timeslice accuracy of 95.6% and a class accuracy of 79.4%.

Modeling interaction in multi-resident activities

“…The design of sequences strongly depends on the type of activities monitored in the environment. Generally, studies consider each day of registration as a sequence when activity data reflects the natural human behavior of residents as in (van Kasteren et al, 2008). But in the case of CASAS "Multi-resident ADLs", many days represent activity data of a single pair of volunteers, while others represent activity data of two pairs of volunteers.…”

Activity recognition in smart homes with self verification of assignments

“…The sensing technology to capture human activity observations is based on either wearable sensors mainly used in physical activities, such as accelerometer and gyroscope [16], or environment interactive sensors used in general activities monitoring, such as light, temperature, motion, pressure and binary contact switch sensors [11]. While the proposed approach is focused more on the general activities category, for a review of the state of the art, we also briefly discuss the existing approaches applied for physical activities.…”

“…In order to exploit the semantic information of domain knowledge, sensor data and activities, context lattices are applied for activity recognition (CL-AR) [37]. HMM is applied and compared with CRF [11] and in order to get a more generalized activity recognition approach, NB, HMM and CRF are compared for activities within a dataset and by combining the common activities of multiple datasets with different environmental settings [30]. HMM requires a large set of training samples and unlike CRF it may not be able to capture long range dependencies of observations [19].…”

“…F1-score closer to 1 shows the best performance, whereas 0 indicates the worst performance. Finally, the Accuracy of the system is measured, which takes into account not only the correctly assigned labels (TPs) but also the activities that are correctly rejected (TNs) out of the total assigned labels Table 1 shows the summary of activities performed in each of the nine publicly available challenging smart home datasets: eight from CASAS [55] and one from Kasteren [11]. Datasets are selected based on the challenges that include the number of participants in each home, overlapping of features among activity instances of different classes, variations in the performance of the same activity instances by different users, addition of noise and errors in the the datasets because of the sensors and the participants, and the presence of non participating agents (pets) in the home that affect the sensor inputs during an activity instance.…”

“…Activity segmentation is performed using different techniques, sliding windows [7], relative weighting of objects in adjacent activities [8] or pattern mining [9], just to name a few. Segmented activity instances are classified in activity classes using different learning models such as Hidden Markov Model (HMM) [10], Conditional Random Fields (CRF) [11], Naive Bayes (NB) [12], Support Vector Machine (SVM) [13], Artificial Neural Network (ANN) [14,15], and Decision Tree (DT) [16]. In activity classification, a false assignment could occur due to the unreliable nature of sensor data [17], incorrect execution of an activity [18], similar activities due to overlapping in features [19] or inability of a learning algorithm to assign the correct label [20].…”