Misha Pavel

The psychophysics of visual search.

Most theories of visual search emphasize issues of limited versus unlimited capacity and serial versus parallel processing. In the present article, we suggest a broader framework based on two principles, one empirical and one theoretical. The empirical principle is to focus on conditions at the intersection of visual search and the simple detection and discrimination paradigms of spatial vision. Such simple search conditions avoid artifacts and phenomena specific to more complex stimuli and tasks. The theoretical principle is to focus on the distinction between high and low threshold theory. While high threshold theory is largely discredited for simple detection and discrimination, it persists in the search literature. Furthermore, a low threshold theory such as signal detection theory can account for some of the phenomena attributed to limited capacity or serial processing. In the body of this article, we compare the predictions of high threshold theory and three versions of signal detection theory to the observed effects of manipulating set size, discriminability, number of targets, response bias, external noise, and distractor heterogeneity. For almost all cases, the results are inconsistent with high threshold theory and are consistent with all three versions of signal detection theory. In the Discussion, these simple theories are generalized to a larger domain that includes search asymmetry, multidimensional judgements including conjunction search, response time, search with multiple eye fixations and more general stimulus conditions. We conclude that low threshold theories can account for simple visual search without invoking mechanisms such as limited capacity or serial processing.

Mobile health technology evaluation: the mHealth evidence workshop.

Creative use of new mobile and wearable health information and sensing technologies (mHealth) has the potential to reduce the cost of health care and improve well-being in numerous ways. These applications are being developed in a variety of domains, but rigorous research is needed to examine the potential, as well as the challenges, of utilizing mobile technologies to improve health outcomes. Currently, evidence is sparse for the efficacy of mHealth. Although these technologies may be appealing and seemingly innocuous, research is needed to assess when, where, and for whom mHealth devices, apps, and systems are efficacious. In order to outline an approach to evidence generation in the field of mHealth that would ensure research is conducted on a rigorous empirical and theoretic foundation, on August 16, 2011, researchers gathered for the mHealth Evidence Workshop at NIH. The current paper presents the results of the workshop. Although the discussions at the meeting were cross-cutting, the areas covered can be categorized broadly into three areas: (1) evaluating assessments; (2) evaluating interventions; and (3) reshaping evidence generation using mHealth. This paper brings these concepts together to describe current evaluation standards, discuss future possibilities, and set a grand goal for the emerging field of mHealth research.

Adjustment Learning and Relevant Component Analysis

We propose a new learning approach for image retrieval, which we call adjustment learning, and demonstrate its use for face recognition and color matching. Our approach is motivated by a frequently encountered problem, namely, that variability in the original data representation which is not relevant to the task may interfere with retrieval and make it very difficult. Our key observation is that in real applications of image retrieval, data sometimes comes in small chunks - small subsets of images that come from the same (but unknown) class. This is the case, for example, when a query is presented via a short video clip. We call these groups chunklets, and we call the paradigm which uses chunklets for unsupervised learning adjustment learning. Within this paradigm we propose a linear scheme, which we call Relevant Component Analysis; this scheme uses the information in such chunklets to reduce irrelevant variability in the data while amplifying relevant variability. We provide results using our method on two problems: face recognition (using a database publicly available on the web), and visual surveillance (using our own data). In the latter application chunklets are obtained automatically from the data without the need of supervision.

Towards ASR on partially corrupted speech

A new highly parallel approach to automatic recognition of speech, inspired by early Fetchers research on articulation index, and based on independent probability estimates in several sub-bands of the available speech spectrum, is presented. The approach is especially suitable for situations when part of the spectrum of speech is computed. In such cases, it can yield an order-of-magnitude improvement in the error rate over a conventional full-band recognizer.

The effect of expectations on slow oculomotor control--IV. Anticipatory smooth eye movements depend on prior target motions.

Prior work had shown that smooth eye movements depend both on the motion of the target on the retina and on the subjects expectations about future target motion (Kowler and Steinman, 1979a,b). Effects of expectation cannot be eliminated by making target motions unpredictable (Kowler and Steinman, 1981). The experiment reported here shows that effects of expectations on smooth eye movement depend in a lawful way on the history of prior target motions. Anticipatory smooth eye movements (involuntary drifts in the direction of future target motion) were measured while subjects fixated a stationary target that was expected to step in an unpredictable direction (right or left). Anticipatory smooth eye movement velocity depended on the sequence of steps in prior trials, e.g. velocity was faster to the right when the prior steps were to the right. The influence of prior steps diminished the further back into the past the step occurred. Sequential dependencies were also observed for the saccades used to track the target steps. Anticipatory smooth eye movement velocity was predicted by a two-state Markov model developed by Falmagne et al. (1975) for similar sequential dependencies observed in a manual reaction-time task (button-pressing). The model uses the prior sequence of target motions to predict the subjects expectation, and assumes that the expectation determines anticipatory smooth eye movement velocity. The fit of the model to the data was good which shows that taking expectations into account is both necessary and feasible. Taking expectations into account, quantitatively, allows accurate predictions about smooth eye movement velocity when target motions are unpredictable.

Intelligent Systems for Assessing Aging Changes: Home-Based, Unobtrusive, and Continuous Assessment of Aging

OBJECTIVES To describe a longitudinal community cohort study, Intelligent Systems for Assessing Aging Changes, that has deployed an unobtrusive home-based assessment platform in many seniors homes in the existing community. METHODS Several types of sensors have been installed in the homes of 265 elderly persons for an average of 33 months. Metrics assessed by the sensors include total daily activity, time out of home, and walking speed. Participants were given a computer as well as training, and computer usage was monitored. Participants are assessed annually with health and function questionnaires, physical examinations, and neuropsychological testing. RESULTS Mean age was 83.3 years, mean years of education was 15.5, and 73% of cohort were women. During a 4-week snapshot, participants left their home twice a day on average for a total of 208 min per day. Mean in-home walking speed was 61.0 cm/s. Participants spent 43% of days on the computer averaging 76 min per day. DISCUSSION These results demonstrate for the first time the feasibility of engaging seniors in a large-scale deployment of in-home activity assessment technology and the successful collection of these activity metrics. We plan to use this platform to determine if continuous unobtrusive monitoring may detect incident cognitive decline.

A Typology for Visualizing Uncertainty

Information analysts must rapidly assess information to determine its usefulness in supporting and informing decision makers. In addition to assessing the content, the analyst must be confident about the quality and veracity of the information. Visualizations can concisely represent vast quantities of information, thus aiding the analyst to examine larger quantities of material; however, visualization programs are challenged to incorporate a notion of confidence or certainty because the factors that influence the certainty or uncertainty of information vary with the type of information and the type of decisions being made. For example, the assessment of potentially subjective human-reported data leads to a large set of uncertainty concerns in fields such as national security, law enforcement (witness reports), and even scientific analysis where data is collected from a variety of individual observers. What’s needed is a formal model or framework for describing uncertainty as it relates to information analysis, to provide a consistent basis for constructing visualizations of uncertainty. This paper proposes an expanded typology for uncertainty, drawing from past frameworks targeted at scientific computing. The typology provides general categories for analytic uncertainty, a framework for creating task-specific refinements to those categories, and examples drawn from the national security field.

Unobtrusive and Ubiquitous In-Home Monitoring: A Methodology for Continuous Assessment of Gait Velocity in Elders

Gait velocity has been shown to quantitatively estimate risk of future hospitalization, a predictor of disability, and has been shown to slow prior to cognitive decline. In this paper, we describe a system for continuous and unobtrusive in-home assessment of gait velocity, a critical metric of function. This system is based on estimating walking speed from noisy time and location data collected by a ¿sensor line¿ of restricted view passive infrared motion detectors. We demonstrate the validity of our system by comparing with measurements from the commercially available GAITRite walkway system gait mat. We present the data from 882 walks from 27 subjects walking at three different subject-paced speeds (encouraged to walk slowly, normal speed, or fast) in two directions through a sensor line. The experimental results show that the uncalibrated system accuracy (average error) of estimated velocity was 7.1 cm/s (SD = 11.3 cm/s), which improved to 1.1 cm/s (SD = 9.1 cm/s) after a simple calibration procedure. Based on the average measured walking speed of 102 cm/s, our system had an average error of less than 7% without calibration and 1.1% with calibration.

On the relative importance of various components of the modulation spectrum for automatic speech recognition

We measured the accuracy of speech recognition as a function of band-pass filtering of the time trajectories of spectral envelopes. We examined (i) several types of recognizers such as dynamic time warping (DTW) and hidden Markov model (HMM), and (ii) several types of features, such as filter bank output, mel-frequency cepstral coefficients (MFCC), and perceptual linear predictive (PLP) coefficients. We used the resulting recognition data to determine the relative importance of information in different modulation spectral components of speech for automatic speech recognition. We concluded that: (1) most of the useful linguistic information is in modulation frequency components from the range between 1 and 16 Hz, with the dominant component at around 4 Hz; (2) in some realistic environments, the use of components from the range below 2 Hz or above 16 Hz can degrade the recognition accuracy.

Advancing the science of mHealth.

Mobile health (mHealth) technologies have the potential to greatly impact health research, health care, and health outcomes, but the exponential growth of the technology has outpaced the science. This article outlines two initiatives designed to enhance the science of mHealth. The mHealth Evidence Workshop used an expert panel to identify optimal methodological approaches for mHealth research. The NIH mHealth Training Institutes address the silos among the many academic and technology areas in mHealth research and is an effort to build the interdisciplinary research capacity of the field. Both address the growing need for high quality mobile health research both in the United States and internationally. mHealth requires a solid, interdisciplinary scientific approach that pairs the rapid change associated with technological progress with a rigorous evaluation approach. The mHealth Evidence Workshop and the NIH mHealth Training Institutes were both designed to address and further develop this scientific approach to mHealth.