Detection and Classification of Whale Acoustic Signals

This dissertation focuses on two vital challenges in relation to whale acoustic signals: detection and classification.

In detection, we evaluated the influence of the uncertain ocean environment on the spectrogram-based detector, and derived the likelihood ratio of the proposed Short Time Fourier Transform detector. Experimental results showed that the proposed detector outperforms detectors based on the spectrogram. The proposed detector is more sensitive to environmental changes because it includes phase information.

In classification, our focus is on finding a robust and sparse representation of whale vocalizations. Because whale vocalizations can be modeled as polynomial phase signals, we can represent the whale calls by their polynomial phase coefficients. In this dissertation, we used the Weyl transform to capture chirp rate information, and used a two dimensional feature set to represent whale vocalizations globally. Experimental results showed that our Weyl feature set outperforms chirplet coefficients and MFCC (Mel Frequency Cepstral Coefficients) when applied to our collected data.

Since whale vocalizations can be represented by polynomial phase coefficients, it is plausible that the signals lie on a manifold parameterized by these coefficients. We also studied the intrinsic structure of high dimensional whale data by exploiting its geometry. Experimental results showed that nonlinear mappings such as Laplacian Eigenmap and ISOMAP outperform linear mappings such as PCA and MDS, suggesting that the whale acoustic data is nonlinear.

We also explored deep learning algorithms on whale acoustic data. We built each layer as convolutions with either a PCA filter bank (PCANet) or a DCT filter bank (DCTNet). With the DCT filter bank, each layer has different a time-frequency scale representation, and from this, one can extract different physical information. Experimental results showed that our PCANet and DCTNet achieve high classification rate on the whale vocalization data set. The word error rate of the DCTNet feature is similar to the MFSC in speech recognition tasks, suggesting that the convolutional network is able to reveal acoustic content of speech signals.

Walking in old age and development of metabolic syndrome: the health, aging, and body composition study.

BACKGROUND: The specific health benefits of meeting physical activity guidelines are unclear in older adults. We examined the association between meeting, not meeting, or change in status of meeting physical activity guidelines through walking and the 5-year incidence of metabolic syndrome in older adults. METHODS: A total of 1,863 Health, Aging, and Body Composition (Health ABC) Study participants aged 70-79 were followed for 5 years (1997-1998 to 2002-2003). Four walking groups were created based on self-report during years 1 and 6: Sustained low (Year 1, <150 min/week, and year 6, <150 min/week), decreased (year 1, >150 min/week, and year 6, <150 min/week), increased (year 1, <150 min/week, and year 6, >150 min/week), and sustained high (year 1, >150 min/week, and year 6, >150 min/week). Based on the Adult Treatment Panel III (ATP III) panel guidelines, the metabolic syndrome criterion was having three of five factors: Large waist circumference, elevated blood pressure, triglycerides, blood glucose, and low high-density lipoprotein (HDL) levels. RESULTS: Compared to the sustained low group, the sustained high group had a 39% reduction in odds of incident metabolic syndrome [adjusted odds ratio (OR) = 0.61; 95% confidence interval (CI), 0.40-0.93], and a significantly lower likelihood of developing the number of metabolic syndrome risk factors that the sustained low group developed over 5 years (beta = -0.16, P = 0.04). CONCLUSIONS: Meeting or exceeding the physical activity guidelines via walking significantly reduced the odds of incident metabolic syndrome and onset of new metabolic syndrome components in older adults. This protective association was found only in individuals who sustained high levels of walking for physical activity.

Impact of Funding Source on Clinical Trial Results Including Cardiovascular Outcome Trials.

Previous authors have suggested a higher likelihood for industry-sponsored (IS) studies to have positive outcomes than non-IS studies, though the influence of publication bias was believed to be a likely confounder. We attempted to control for the latter using a prepublication database to compare the primary outcome of recent trials based on sponsorship. We used the "advanced search" feature in the clinicaltrials.gov website to identify recently completed phase III studies involving the implementation of a pharmaceutical agent or device for which primary data were available. Studies were categorized as either National Institutes of Health (NIH) sponsored or IS. Results were labeled "favorable" if the results favored the intervention under investigation or "unfavorable" if the intervention fared worse than standard medical treatment. We also performed an independent literature search to identify the cardiovascular trials as a case example and again categorized them into IS versus NIH sponsored. A total of 226 studies sponsored by NIH were found. When these were compared with the latest 226 IS studies, it was found that IS studies were almost 4 times more likely to report a positive outcome (odds ratio [OR] 3.90, 95% confidence interval [CI] 2.6087 to 5.9680, p <0.0001). As a case example of a specialty, we also identified 25 NIH-sponsored and 215 IS cardiovascular trials, with most focusing on hypertension therapy (31.6%) and anticoagulation (17.9%). IS studies were 7 times more likely to report favorable outcomes (OR 7.54, 95% CI 2.19 to 25.94, p = 0.0014). They were also considerably less likely to report unfavorable outcomes (OR 0.11, 95% CI 0.04 to 0.26, p <0.0001). In conclusion, the outcomes of large clinical studies especially cardiovascular differ considerably on the basis of their funding source, and publication bias appears to have limited influence on these findings.