Balakumar Balasingam

Sooraj Sunil, Balakumar Balasingam

The rapid adoption of deep learning has increasingly led to data-driven models replacing classical model-based algorithms, even in domains governed by well-understood physical laws. While data-driven models, such as long short-term memory (LSTM) networks, have become a popular choice for time-series analysis, their performance relative to model-based approaches in structured environments is rarely evaluated objectively. This paper presents a performance evaluation framework comparing an LSTM classifier against a model-based expectation maximization (EM) classifier for binary time-series classification. The evaluation is conducted on two scalar linear Gaussian state space models differing only in their noise statistics, where the Kalman filter likelihood ratio test with true parameters serves as a reference for the best achievable classification performance.Through Monte Carlo simulations, the classifiers are evaluated across three axes: task difficulty, controlled by the separation in process or measurement noise between the two models; sequence length; and training dataset size. The results show that the EM classifier, which exploits the known model structure, performs strongly when the data conform to the assumed model class. The LSTM classifier requires a larger separation in noise statistics to achieve reliable classification, and its performance saturates below the reference classifier when the models differ only in measurement noise, regardless of sequence length or training dataset size.

Stephen Bottos, Balakumar Balasingam

In this paper, we propose an approach to track the progression of eye-gaze while reading a block of text on computer screen. The proposed approach will help to accurately quantify reading, e.g., identifying the lines of text that were read/skipped and estimating the time spent on each line, based on commercially available inexpensive eye-tracking devices. The proposed approach is based on a novel Slip Kalman filter that is custom designed to track the progression of reading. The performance of the proposed method is demonstrated using 25 pages eye-tracking data collected using a commercial desk-mounted eye-tracking device.

Stephen Bottos, Balakumar Balasingam

In this paper we consider the problem of tracking the progression of reading through eye-gaze measurements. Such an algorithm is novel and will ultimately help to develop a method of analyzing eye-gaze data which had been collected during reading activity in order to uncover crucial information regarding the individual's interest level and quality of experience while reading a passage of text or book. Additionally, such an approach will serve as a "visual signature" - a method of verifying if an individual has indeed given adequate attention to critical text-based information. Further, an accurate "reading-progression-tracker" has potential applications in educational institutions, e-readers and parenting solutions. Tracking the progression of reading remains a challenging problem due to the fact that eye-gaze movements are highly noisy and the eye-gaze is easily distracted in a limited space, like an e-book. In a prior work, we proposed an approach to analyze eye-gaze fixation points collected while reading a page of text in order to classify each measurement to a line of text; this approach did not consider tracking the progression of reading along the line of text. In this paper, we extend the capabilities of the previous algorithm in order to accurately track the progression of reading along each line. the proposed approach employs least squares batch estimation in order to estimate three states of the horizontal saccade: position, velocity and acceleration. First, the proposed approach is objectively evaluated on a simulated eye-gaze dataset. Then, the proposed algorithm is demonstrated on real data collected by a Gazepoint eye-tracker while the subject is reading several pages from an electronic book.

Stephen Bottos, Balakumar Balasingam

In this paper, we consider the problem of tracking the eye-gaze of individuals while they engage in reading. Particularly, we develop ways to accurately track the line being read by an individual using commercially available eye tracking devices. Such an approach will enable futuristic functionalities such as comprehension evaluation, interest level detection, and user-assisting applications like hands-free navigation and automatic scrolling. Existing commercial eye trackers provide an estimated location of the eye-gaze fixations every few milliseconds. However, this estimated data is found to be very noisy. As such, commercial eye-trackers are unable to accurately track lines while reading. In this paper we propose several statistical models to bridge the commercial gaze tracker outputs and eye-gaze patterns while reading. We then employ hidden Markov models to parametrize these statistical models and to accurately detect the line being read. The proposed approach is shown to yield an improvement of over 20% in line detection accuracy.