Martin L. Griss

Activity-Aware Mental Stress Detection Using Physiological Sensors

Continuous stress monitoring may help users better under- stand their stress patterns and provide physicians with more reliable data for interventions. Previously, studies on mental stress detection were lim- ited to a laboratory environment where participants generally rested in a sedentary position. However, it is impractical to exclude the effects of physical activity while developing a pervasive stress monitoring appli- cation for everyday use. The physiological responses caused by mental stress can be masked by variations due to physical activity.

NuActiv: recognizing unseen new activities using semantic attribute-based learning

We study the problem of how to recognize a new human activity when we have never seen any training example of that activity before. Recognizing human activities is an essential element for user-centric and context-aware applications. Previous studies showed promising results using various machine learning algorithms. However, most existing methods can only recognize the activities that were previously seen in the training data. A previously unseen activity class cannot be recognized if there were no training samples in the dataset. Even if all of the activities can be enumerated in advance, labeled samples are often time consuming and expensive to get, as they require huge effort from human annotators or experts. In this paper, we present NuActiv, an activity recognition system that can recognize a human activity even when there are no training data for that activity class. Firstly, we designed a new representation of activities using semantic attributes, where each attribute is a human readable term that describes a basic element or an inherent characteristic of an activity. Secondly, based on this representation, a two-layer zero-shot learning algorithm is developed for activity recognition. Finally, to reinforce recognition accuracy using minimal user feedback, we developed an active learning algorithm for activity recognition. Our approach is evaluated on two datasets, including a 10-exercise-activity dataset we collected, and a public dataset of 34 daily life activities. Experimental results show that using semantic attribute-based learning, NuActiv can generalize knowledge to recognize unseen new activities. Our approach achieved up to 79% accuracy in unseen activity recognition.

WASP: an enhanced indoor locationing algorithm for a congested Wi-Fi environment

Accurate and reliable location information is important to many context-aware mobile applications. While the Global Positioning System (GPS) works quite well outside, it is quite problematic for indoor locationing. In this paper, we introduce WASP, an enhanced indoor locationing algorithm. WASP is based on the Redpin algorithm which matches the received Wi-Fi signal with the signals in the training data and uses the position of the closest training data as the users current location. However, in a congested Wi-Fi environment the Redpin algorithm gets confused because of the unstable radio signals received from too many APs. WASP addresses this issue by voting the right location from more neighboring training examples, weighting Access Points (AP) based on their correlation with a certain location, and automatic filtering of noisy APs. WASP significantly outperform the-state-of-the-art Redpin algorithm. In addition, this paper also reports our findings on how the size of the training data, the physical size of the room and the number of APs affect the accuracy of indoor locationing.

n-gram geo-trace modeling

As location-sensing smart phones and location-based services gain mainstream popularity, there is increased interest in developing techniques that can detect anomalous activities. Anomaly detection capabilities can be used in theft detection, remote elder-care monitoring systems, and many other applications. In this paper we present an n-gram based model for modeling a users mobility patterns. Under the Markovian assumption that a users location at time t depends only on the last n - 1 locations until t - 1, we can model a users idiosyncratic location patterns through a collection of n-gram geo-labels, each with estimated probabilities. We present extensive evaluations of the n-gram model conducted on real-world data, compare it with the previous approaches of using T-Patterns and Markovian models, and show that for anomaly detection the n-gram model outperforms existing work by approximately 10%. We also show that the model can use a hierarchical location partitioning system that is able to obscure a users exact location, to protect privacy, while still allowing applications to utilize the obscured location data for modeling anomalies effectively.

Nonparametric discovery of human routines from sensor data

People engage in routine behaviors. Automatic routine discovery goes beyond low-level activity recognition such as sitting or standing and analyzes human behaviors at a higher level (e.g., commuting to work). With recent developments in ubiquitous sensor technologies, it becomes easier to acquire a massive amount of sensor data. One main line of research is to mine human routines from sensor data using parametric topic models such as latent Dirichlet allocation. The main shortcoming of parametric models is that it assumes a fixed, pre-specified parameter regardless of the data. Choosing an appropriate parameter usually requires an inefficient trial-and-error model selection process. Furthermore, it is even more difficult to find optimal parameter values in advance for personalized applications. In this paper, we present a novel nonparametric framework for human routine discovery that can infer high-level routines without knowing the number of latent topics beforehand. Our approach is evaluated on public datasets in two routine domains: a 34-daily-activity dataset and a transportation mode dataset. Experimental results show that our nonparametric framework can automatically learn the appropriate model parameters from sensor data without any form of model selection procedure and can outperform traditional parametric approaches for human routine discovery tasks.

Overseer: A Mobile Context-Aware Collaboration and Task Management System for Disaster Response

Efficient collaboration and task management is challenging in distributed, dynamically-formed organizations such as ad hoc disaster response teams. Ineffective collaboration may result in poor performance and loss of life. In this paper, we present Overseer, an open multi-agent system that leverages context information in a mobile setting to facilitate collaboration and task allocation. We describe our system architecture, deployment, evaluation metrics, challenges and proposed solutions. We also show how mobile context can be used to create dynamic role-based assignments to support collaboration and effective task management.

Sensor Data as a Service -- A Federated Platform for Mobile Data-centric Service Development and Sharing

The Internet of Things (IoT) offers the promise of integrating the digital world of the Internet with the physical world in which we live. But realizing this promise necessitates a systematic approach to integrating the sensors, actuators, and information on which they operate into the Internet we know today. This paper reports the design and development of an open community-oriented platform aiming to support federated sensor data as a service, featuring interoperability and reusability of heterogeneous sensor data and data services. The concepts of virtual sensors and virtual devices are identified as central autonomic units to model scalable and context-aware configurable/reconfigurable sensor data and services. The decoupling of the storage and management of sensor data and platform-oriented metadata enables the handling of both discrete and streaming sensor data. A cloud computing-empowered prototyping system has been established as a proof of concept to host smart community-oriented sensor data and services.

Towards zero-shot learning for human activity recognition using semantic attribute sequence model

Understanding human activities is important for user-centric and context-aware applications. Previous studies showed promising results using various machine learning algorithms. However, most existing methods can only recognize the activities that were previously seen in the training data. In this paper, we present a new zero-shot learning framework for human activity recognition that can recognize an unseen new activity even when there are no training samples of that activity in the dataset. We propose a semantic attribute sequence model that takes into account both the hierarchical and sequential nature of activity data. Evaluation on datasets in two activity domains show that the proposed zero-shot learning approach achieves 70-75% precision and recall recognizing unseen new activities, and outperforms supervised learning with limited labeled data for the new classes.

A story-centered, learn-by-doing approach to software engineering education

Carnegie Mellons West Coast Campus offers an MS in Software Engineering, with technical and development management tracks, targeted at working software professionals in Silicon Valley. We believe the program to be unique in that it is entirely team-based and project-centered. Students learn by doing as they are coached just in time by faculty in the context of authentic projects, and they are evaluated based on what they produce. Student satisfaction is high: 92% believe that the program has given them a competitive advantage with respect to their professional peers, and their promotion and salary histories bear out this belief.

Hermes -- A Context-Aware Application Development Framework and Toolkit for the Mobile Environment

The modern mobile ubiquitous computing environment, with reasonable connectivity, processing power and sensing capabilities on portable devices, present applications and services with the opportunity to be truly context-aware. But building context-aware applications has large development overheads due to complexities of sensing, aggregating and inferencing context information. To reduce the developers burden, over the last decade, many context-aware application development frameworks and toolkits have been created. Yet none of these existing solutions address the specific challenges, opportunities and constraints presented by the modern mobile computing environment. This paper describes a next generation software toolkit which provides a framework, including context inferencing, communication, storage, power management, security and intelligibility support, for developing more powerful context-aware applications for the modern mobile environment. The Hermes toolkit and framework are designed around a loosely coupled component-based architecture that facilitates the decomposition of context-aware applications into multiple smaller components, each of which captures, transforms or aggregates pieces of context information to produce the high-level context used by applications. In this paper we present the overall architecture of Hermes, a sample application design and our implementation details.