Panu Korpipää

Activity classification using realistic data from wearable sensors

Automatic classification of everyday activities can be used for promotion of health-enhancing physical activities and a healthier lifestyle. In this paper, methods used for classification of everyday activities like walking, running, and cycling are described. The aim of the study was to find out how to recognize activities, which sensors are useful and what kind of signal processing and classification is required. A large and realistic data library of sensor data was collected. Sixteen test persons took part in the data collection, resulting in approximately 31 h of annotated, 35-channel data recorded in an everyday environment. The test persons carried a set of wearable sensors while performing several activities during the 2-h measurement session. Classification results of three classifiers are shown: custom decision tree, automatically generated decision tree, and artificial neural network. The classification accuracies using leave-one-subject-out cross validation range from 58 to 97% for custom decision tree classifier, from 56 to 97% for automatically generated decision tree, and from 22 to 96% for artificial neural network. Total classification accuracy is 82% for custom decision tree classifier, 86% for automatically generated decision tree, and 82% for artificial neural network

Managing context information in mobile devices

We present a uniform mobile terminal software framework that provides systematic methods for acquiring and processing useful context information from a users surroundings and giving it to applications. The framework simplifies the development of context-aware mobile applications by managing raw context information gained from multiple sources and enabling higher-level context abstractions.

Accelerometer-based gesture control for a design environment

Accelerometer-based gesture control is studied as a supplementary or an alternative interaction modality. Gesture commands freely trainable by the user can be used for controlling external devices with handheld wireless sensor unit. Two user studies are presented. The first study concerns finding gestures for controlling a design environment (Smart Design Studio), TV, VCR, and lighting. The results indicate that different people usually prefer different gestures for the same task, and hence it should be possible to personalise them. The second user study concerns evaluating the usefulness of the gesture modality compared to other interaction modalities for controlling a design environment. The other modalities were speech, RFID-based physical tangible objects, laser-tracked pen, and PDA stylus. The results suggest that gestures are a natural modality for certain tasks, and can augment other modalities. Gesture commands were found to be natural, especially for commands with spatial association in design environment control.

Bayesian approach to sensor-based context awareness

AbstractThe usability of a mobile device and services can be enhanced by context awareness. The aim of this experiment was to expand the set of generally recognizable constituents of context concerning personal mobile device usage. Naive Bayesian networks were applied to classify the contexts of a mobile device user in her normal daily activities. The distinguishing feature of this experiment in comparison to earlier context recognition research is the use of a naive Bayes framework, and an extensive set of audio features derived partly from the algorithms of the upcoming MPEG-7 standard. The classification was based mainly on audio features measured in a home scenario. The classification results indicate that with a resolution of one second in segments of 5–30 seconds, situations can be extracted fairly well, but most of the contexts are likely to be valid only in a restricted scenario. Naive Bayes framework is feasible for context recognition. In real world conditions, the recognition accuracy using leave-one-out cross validation was 87% of true positives and 95% of true negatives, averaged over nine eight-minute scenarios containing 17 segments of different lengths and nine different contexts. Respectively, the reference accuracies measured by testing with training data were 88% and 95%, suggesting that the model was capable of covering the variability introduced in the data on purpose. Reference recognition accuracy in controlled conditions was 96% and 100%, respectively. However, from the applicability viewpoint, generalization remains a problem, as from a wider perspective almost any feature may refer to many possible real world situations.

An ontology for mobile device sensor-based context awareness

In mobile computing, the efficient utilisation of the information gained from the sensors embedded in the devices is difficult. Instead of using raw measurement data application specifically, as currently is customary, higher abstraction level semantic descriptions of the situation, context, can be used to develop mobile applications that are more usable. This article introduces an ontology of context constituents, which are derived from a set of sensors embedded in a mobile device. In other words, a semantic interface to the sensor data is provided. The ontology promotes the rapid development of mobile applications, more efficient use of resources, as well as reuse and sharing of information between communicating entities. A few mobile applications are presented to illustrate the possibilities of using the ontology.

Utilising context ontology in mobile device application personalisation

Context Studio, an application personalisation tool for semi-automated context-based adaptation, has been proposed to provide a flexible means of implementing context-aware features. In this paper, Context Studio is further developed for the end users of small-screen mobile devices. Navigating and information presentation are designed for small screens, especially for the Series 60 mobile phone user interface. Context ontology, with an enhanced vocabulary model, is utilized to offer scalable representation and easy navigation of context and action information in the UI. The ontology vocabulary hierarchy is transformed into a folder-file model representation in the graphical user interface. UI elements can be directly updated, according to the extensions and modifications to ontology vocabularies, automatically in an online system. A rule model is utilized to allow systematic management and presentation of context-action rules in the user interface. The chosen ontology-based UI model is evaluated with a usability study.

USER INDEPENDENT GESTURE INTERACTION FOR SMALL HANDHELD DEVICES

Accelerometer-based gesture recognition facilitates a complementary interaction modality for controlling mobile devices and home appliances. Using gestures for the task of home appliance control requires use of the same device and gestures by different persons, i.e. user independent gesture recognition. The practical application in small embedded low-resource devices also requires high computational performance. The user independent gesture recognition accuracy was evaluated with a set of eight gestures and seven users, with a total of 1120 gestures in the dataset. Twenty-state continuous HMM yielded an average of 96.9% user independent recognition accuracy, which was cross-validated by leaving one user in turn out of the training set. Continuous and discrete five-state HMM computational performances were compared with a reference test in a PC environment, indicating that discrete HMM is 20% faster. Computational performance of discrete five-state HMM was evaluated in an embedded hardware environment with a 104 MHz ARM-9 processor and Symbian OS. The average recognition time per gesture calculated from 1120 gesture repetitions was 8.3 ms. With this result, the computational performance difference between the compared methods is considered insignificant in terms of practical application. Continuous HMM is hence recommended as a preferred method due to its better suitability for a continuous-valued signal, and better recognition accuracy. The results suggest that, according to both evaluation criteria, HMM is feasible for practical user independent gesture control applications in mobile low-resource embedded environments.

Using PCA and ICA for exploratory data analysis in situation awareness

This paper presents an approach for analyzing hand held device usage situation (context) phenomena. The situation information under examination is multidimensional fuzzy feature information derived from multisensor measurements. The analysis is conducted using principal component analysis (PCA) and independent component analysis (ICA). PCA is used to fuse multidimensional feature information into a more compact representation while the ICA is applied to extract patterns containing independent low level information about the situation. The results show that a few principal components compress the situation data representation efficiently. In addition, principal component representation provides a method for visualizing high level situation information. Most independent components extracted from the usage situation data correlate strongly with some of the original signals. This suggests that the original context data already consist of relatively independent signals if the temporal relations in the data are omitted.

Interaction and end-user programming with a context-aware mobile application

In this paper we present the user interface design and evaluation of a tool for customizing mobile phone applications with context-aware features. The tool enables the user to program a set of context-action rules, defining the behavior of the device when a selected context is recognized and/or some other user-defined conditions are met. The tool user interface design is described starting from an early paper prototype and its evaluation, leading to a functional software implementation in a mobile phone. Finally, the usability evaluation of the functional prototype, and other relevant findings from the user test, are presented.

Extracting the Context of a Mobile Device User

Abstract It has become increasingly important in mobile computing to be able to recognize the situation i.e. the context of a mobile device user, in order to enhance the effectiveness of human computer interaction. This paper describes an approach to the extraction of higher level contexts from the multidimensional low level context information, focusing on the analysis and comparison of clustering and segmentation behaviour of crisp versus fuzzy information. Context clustering is performed by using the k-means algorithm and segmentation by using a minimum-variance algorithm. The results indicate that fuzzy quantization is superior to the crisp quantization with k-means, yielding more consistent clustering. Segment borders found from the fuzzy data correspond better to the real context changes than those found from the crisp data.