Jó Ágila Bitsch

EEG-based automatic emotion recognition: Feature extraction, selection and classification methods

Automatic emotion recognition is an interdisciplinary research field which deals with the algorithmic detection of human affect, e.g. anger or sadness, from a variety of sources, such as speech or facial gestures. Apart from the obvious usage for industry applications in human-robot interaction, acquiring the emotional state of a person automatically also is of great potential for the health domain, especially in psychology and psychiatry. Here, evaluation of human emotion is often done using oral feedback or questionnaires during doctor-patient sessions. However, this can be perceived as intrusive by the patient. Furthermore, the evaluation can only be done in a noncontinuous manner, e.g. once a week during therapy sessions. In contrast, using automatic emotion detection, the affect state of a person can be evaluated in a continuous non-intrusive manner, for example to detect early on-sets of depression. An additional benefit of automatic emotion recognition is the objectivity of such an approach, which is not influenced by the perception of the patient and the doctor. To reach the goal of objectivity, it is important, that the source of the emotion is not easily manipulable, e.g. as in the speech modality. To circumvent this caveat, novel approaches in emotion detection research the potential of using physiological measures, such as galvanic skin sensors or pulse meters. In this paper we outline a way of detecting emotion from brain waves, i.e., EEG data. While EEG allows for a continuous, real-time automatic emotion recognition, it furthermore has the charm of measuring the affect close to the point of emergence: the brain. Using EEG data for emotion detection is nevertheless a challenging task: Which features, EEG channel locations and frequency bands are best suited for is an issue of ongoing research. In this paper we evaluate the use of state of the art feature extraction, feature selection and classification algorithms for EEG emotion classification using data from the de facto standard dataset, DEAP. Moreover, we present results that help choose methods to enhance classification performance while simultaneously reducing computational complexity.

Ratpack: Wearable sensor networks for animal observation

The goal of our project is to describe the behavior of rats. For this purpose we are using wireless sensor networks, monitoring various quantities that yield important information to complement current knowledge on the behavioral repertoire of rats. So far, on the sensing and processing side we have developed innovative, minimalist approaches pointing in two directions: vocalization analysis and movement tracking. On the data collection and routing side we have adapted to the known burrowing habits of rats by developing new methods for synchronization and data aggregation under the paradigm of sporadic connectivity in a sparse, dynamic network.

Smart sensors for small rodent observation

Working towards the observation of rats (and other small rodents) in the wild we have developed tools that will enable us to study their behavior using a wireless network of wearable sensor nodes. The space and weight constraints resulting from the size of the animals have led to simple but functional approaches for vocalization classification and position estimation. For the resulting data we have developed novel, delay-tolerant routing and collection strategies. These are expected to be used in a sparse, dynamic network resulting from various rats being tagged with our nodes and running around freely - an area that will eventually be too big to be covered solely by stationary data sinks. Furthermore, the system is designed to extract information on the social interactions between animals from the routing data. It currently works in an indoor environment and we are preparing it for tests in a controlled outdoor setup.

Psychologist in a Pocket: Lexicon Development and Content Validation of a Mobile-Based App for Depression Screening.

Background Language reflects the state of one’s mental health and personal characteristics. It also reveals preoccupations with a particular schema, thus possibly providing insights into psychological conditions. Using text or lexical analysis in exploring depression, negative schemas and self-focusing tendencies may be depicted. As mobile technology has become highly integrated in daily routine, mobile devices have the capacity for ecological momentary assessment (EMA), specifically the experience sampling method (ESM), where behavior is captured in real-time or closer in time to experience in one’s natural environment. Extending mobile technology to psychological health could augment initial clinical assessment, particularly of mood disturbances, such as depression and analyze daily activities, such as language use in communication. Here, we present the process of lexicon generation and development and the initial validation of Psychologist in a Pocket (PiaP), a mobile app designed to screen signs of depression through text analysis. Objective The main objectives of the study are (1) to generate and develop a depressive lexicon that can be used for screening text-input in mobile apps to be used in the PiaP; and (2) to conduct content validation as initial validation. Methods The first phase of our research focused on lexicon development. Words related to depression and its symptoms based on the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) and in the ICD-10 Classification of Mental and Behavioural Disorders: Clinical Descriptions and Diagnostic Guidelines classification systems were gathered from focus group discussions with Filipino college students, interviews with mental health professionals, and the review of established scales for depression and other related constructs. Results The lexicon development phase yielded a database consisting of 13 categories based on the criteria depressive symptoms in the DSM-5 and ICD-10. For the draft of the depression lexicon for PiaP, we were able to gather 1762 main keywords and 9655 derivatives of main keywords. In addition, we compiled 823,869 spelling variations. Keywords included negatively-valenced words like “sad”, “unworthy”, or “tired” which are almost always accompanied by personal pronouns, such as “I”, “I’m” or “my” and in Filipino, “ako” or “ko”. For the content validation, only keywords with CVR equal to or more than 0.75 were included in the depression lexicon test-run version. The mean of all CVRs yielded a high overall CVI of 0.90. A total of 1498 main keywords, 8911 derivatives of main keywords, and 783,140 spelling variations, with a total of 793, 553 keywords now comprise the test-run version. Conclusions The generation of the depression lexicon is relatively exhaustive. The breadth of keywords used in text analysis incorporates the characteristic expressions of depression and its related constructs by a particular culture and age group. A content-validated mobile health app, PiaP may help augment a more effective and early detection of depressive symptoms.

The mobile sleep lab app: An open-source framework for mobile sleep assessment based on consumer-grade wearable devices

BACKGROUND Sleep disorders have a prevalence of up to 50% and are commonly diagnosed using polysomnography. However, polysomnography requires trained staff and specific equipment in a laboratory setting, which are expensive and limited resources are available. Mobile and wearable devices such as fitness wristbands can perform limited sleep monitoring but are not evaluated well. Here, the development and evaluation of a mobile application to record and synchronize data from consumer-grade sensors suitable for sleep monitoring is presented and evaluated for data collection capability in a clinical trial. METHODS Wearable and ambient consumer-grade sensors were selected to mimic the functionalities of clinical sleep laboratories. Then, a modular application was developed for recording, processing and visualizing the sensor data. A validation was performed in three phases: (1) sensor functionalities were evaluated, (2) self-experiments were performed in full-night experiments, and (3) the application was tested for usability in a clinical trial on primary snoring. RESULTS The evaluation of the sensors indicated their suitability for assessing basic sleep characteristics. Additionally, the application successfully recorded full-night sleep. The collected data was of sufficient quality to detect and measure body movements, cardiac activity, snoring and brightness. The ongoing clinical trial phase showed the successful deployment of the application by medical professionals. CONCLUSION The proposed software demonstrated a strong potential for medical usage. With low costs, it can be proposed for screening, long-term monitoring or in resource-austere environments. However, further validations are needed, in particular the comparison to a clinical sleep laboratory.

Networking Wireless Energy in Embedded Networks

Wireless energy transfer has recently emerged as a promising alternative to realize the vision of perpetual embedded sensing. However, this technology transforms the notion of energy from merely a node’s local commodity to, similarly to data, a deployment-wide shareable resource. The challenges of managing a shareable energy resource are much more complicated and radically different from the research of the past decade: Besides energy-efficient operation of individual devices, we also need to optimize networkwide energy distribution. To counteract these challenges, we propose an energy stack, a layered software model for energy management in future transiently powered embedded networks. An initial specification of the energy stack, which is based on the historically successful layered approach for data networking, consists of three layers: (i) the transfer layer, which deals with the physical transfer of energy; (ii) the scheduling layer, which optimizes energy distribution over a single hop; and (iii) the network layer, creates a global view of the energy in the network for optimizing its networkwide distribution. As a contribution, we define the interfacing APIs between these layers, delineate their responsibilities, identify corresponding challenges, and provide a first implementation of the energy stack. Our evaluation, using both experimental deployments and high-level simulations, establishes the feasibility of a layered solution to energy management under transient power.

Incremental Checkpointing for Interruptible Computations: Poster Abstract

We propose incremental checkpointing techniques enabling transiently powered devices to retain computational state across multiple activation cycles. As opposed to the existing approaches, which checkpoint complete program state, the proposed techniques keep track of modified RAM locations to incrementally update the retained state in secondary memory, significantly reducing checkpointing overhead both in terms of time and energy.