scholarly journals Machine Source Localization of Tursiops truncatus Whistle-like Sounds in a Reverberant Aquatic Environment

2019 ◽  
Author(s):  
SF Woodward ◽  
D Reiss ◽  
MO Magnasco
Keyword(s):  
Source Localization ◽  
Tursiops Truncatus ◽  
Classification Tree ◽  
Bottlenose Dolphins ◽  
Time Difference ◽  
Process Models ◽  
Training Data ◽  
Support Vector ◽  
Linear Discriminant ◽  
Signature Whistles

AbstractMost research into bottlenose dolphins’ (Tursiops truncatus’) capacity for communication has centered on tonal calls termed whistles, in particular individually distinctive contact calls referred to as signature whistles. While “non-signature” whistles exist, and may be important components of bottlenose dolphins’ communicative repertoire, they have not been studied extensively. This is in part due to the difficulty of attributing whistles to specific individuals, a challenge that has limited the study of not only non-signature whistles but the study of general acoustic exchanges among socializing dolphins. In this paper, we propose the first machine-learning-based approach to identifying the source locations of semi-stationary, tonal, whistle-like sounds in a highly reverberant space, specifically a half-cylindrical dolphin pool. We deliver estimated time-difference-of-arrivals (TDOA’s) and normalized cross-correlation values computed from pairs of hydrophone signals to a random forest model for high-feature-volume classification and feature selection, and subsequently deliver the selected features into linear discriminant analysis, linear and quadratic Support Vector Machine (SVM), and Gaussian process models. In our 14-source-location setup, we achieve perfect accuracy in localization by classification and high accuracy in localization by regression (median absolute deviation of 0.66 m, interquartile range of 0.34 m - 1.57 m), with fewer than 10,000 features. By building a parsimonious (minimum-feature) classification tree model for the same task, we show that a minimally sufficient feature set is consistent with the information valued by a strictly geometric, time-difference-of-arrival-based approach to sound source localization. Ultimately, our regression models yielded better accuracy than the established Steered-Response Power (SRP) method when all training data were used, and comparable accuracy along the pool surface when deprived of training data at testing sites; our methods additionally boast improved computation time and the potential for superior localization accuracy in all dimensions with more training data.

scholarly journals Mie scattering and microparticle-based characterization of heavy metal ions and classification by statistical inference methods

2019 ◽  
Vol 6 (5) ◽  
pp. 190001 ◽  
Author(s):  
Katherine E. Klug ◽  
Christian M. Jennings ◽  
Nicholas Lytal ◽  
Lingling An ◽  
Jeong-Yeol Yoon
Keyword(s):  
Heavy Metal ◽  
Support Vector Machine ◽  
Metal Ions ◽  
Heavy Metal Ions ◽  
Mie Scattering ◽  
Training Data ◽  
Scattering Data ◽  
Support Vector ◽  
Linear Discriminant ◽  
Machine Analysis

A straightforward method for classifying heavy metal ions in water is proposed using statistical classification and clustering techniques from non-specific microparticle scattering data. A set of carboxylated polystyrene microparticles of sizes 0.91, 0.75 and 0.40 µm was mixed with the solutions of nine heavy metal ions and two control cations, and scattering measurements were collected at two angles optimized for scattering from non-aggregated and aggregated particles. Classification of these observations was conducted and compared among several machine learning techniques, including linear discriminant analysis, support vector machine analysis, K-means clustering and K-medians clustering. This study found the highest classification accuracy using the linear discriminant and support vector machine analysis, each reporting high classification rates for heavy metal ions with respect to the model. This may be attributed to moderate correlation between detection angle and particle size. These classification models provide reasonable discrimination between most ion species, with the highest distinction seen for Pb(II), Cd(II), Ni(II) and Co(II), followed by Fe(II) and Fe(III), potentially due to its known sorption with carboxyl groups. The support vector machine analysis was also applied to three different mixture solutions representing leaching from pipes and mine tailings, and showed good correlation with single-species data, specifically with Pb(II) and Ni(II). With more expansive training data and further processing, this method shows promise for low-cost and portable heavy metal identification and sensing.

Mark–recapture of individually distinctive calls—a case study with signature whistles of bottlenose dolphins (Tursiops truncatus)

2020 ◽  
Vol 101 (5) ◽  
pp. 1289-1301
Author(s):  
Emma G Longden ◽  
Simon H Elwen ◽  
Barry McGovern ◽  
Bridget S James ◽  
Clare B Embling ◽  
...  
Keyword(s):  
Tursiops Truncatus ◽  
Bottlenose Dolphins ◽  
Weather Conditions ◽  
Acoustic Monitoring ◽  
Initial Increase ◽  
Visual Methods ◽  
Mark Recapture ◽  
Animal Populations ◽  
Signature Whistles ◽  
Using Data

Abstract Robust abundance estimates of wild animal populations are needed to inform management policies and are often obtained through mark–recapture (MR) studies. Visual methods are commonly used, which limits data collection to daylight hours and good weather conditions. Passive acoustic monitoring offers an alternative, particularly if acoustic cues are naturally produced and individually distinctive. Here we investigate the potential of using individually distinctive signature whistles in a MR framework and evaluate different components of study design. We analyzed signature whistles of common bottlenose dolphins, Tursiops truncatus, using data collected from static acoustic monitoring devices deployed in Walvis Bay, Namibia. Signature whistle types (SWTs) were identified using a bout analysis approach (SIGnature IDentification [SIGID]—Janik et al. 2013). We investigated spatial variation in capture by comparing 21 synchronized recording days across four sites, and temporal variation from 125 recording days at one high-use site (Aphrodite Beach). Despite dolphin vocalizations (i.e., echolocation clicks) being detected at each site, SWTs were not detected at all sites and there was high variability in capture rates among sites where SWTs were detected (range 0–21 SWTs detected). At Aphrodite Beach, 53 SWTs were captured over 6 months and discovery curves showed an initial increase in newly detected SWTs, approaching asymptote during the fourth month. A Huggins closed capture model constructed from SWT capture histories at Aphrodite Beach estimated a population of 54–68 individuals from acoustic detection, which overlaps with the known population size (54–76 individuals—Elwen et al. 2019). This study demonstrates the potential power of using signature whistles as proxies for individual occurrence and in MR abundance estimation, but also highlights challenges in using this approach.

scholarly journals Detection and Characterization of Physical Activity and Psychological Stress from Wristband Data

Signals ◽  
2020 ◽  
Vol 1 (2) ◽  
pp. 188-208
Author(s):  
Mert Sevil ◽  
Mudassir Rashid ◽  
Mohammad Reza Askari ◽  
Zacharie Maloney ◽  
Iman Hajizadeh ◽  
...  
Keyword(s):  
Physical Activity ◽  
Signal Processing ◽  
Feature Extraction ◽  
Psychological Stress ◽  
Training Data ◽  
Support Vector ◽  
K Nearest Neighbor ◽  
Data Set ◽  
Linear Discriminant ◽  
Physiological Variables

Wearable devices continuously measure multiple physiological variables to inform users of health and behavior indicators. The computed health indicators must rely on informative signals obtained by processing the raw physiological variables with powerful noise- and artifacts-filtering algorithms. In this study, we aimed to elucidate the effects of signal processing techniques on the accuracy of detecting and discriminating physical activity (PA) and acute psychological stress (APS) using physiological measurements (blood volume pulse, heart rate, skin temperature, galvanic skin response, and accelerometer) collected from a wristband. Data from 207 experiments involving 24 subjects were used to develop signal processing, feature extraction, and machine learning (ML) algorithms that can detect and discriminate PA and APS when they occur individually or concurrently, classify different types of PA and APS, and estimate energy expenditure (EE). Training data were used to generate feature variables from the physiological variables and develop ML models (naïve Bayes, decision tree, k-nearest neighbor, linear discriminant, ensemble learning, and support vector machine). Results from an independent labeled testing data set demonstrate that PA was detected and classified with an accuracy of 99.3%, and APS was detected and classified with an accuracy of 92.7%, whereas the simultaneous occurrences of both PA and APS were detected and classified with an accuracy of 89.9% (relative to actual class labels), and EE was estimated with a low mean absolute error of 0.02 metabolic equivalent of task (MET).The data filtering and adaptive noise cancellation techniques used to mitigate the effects of noise and artifacts on the classification results increased the detection and discrimination accuracy by 0.7% and 3.0% for PA and APS, respectively, and by 18% for EE estimation. The results demonstrate the physiological measurements from wristband devices are susceptible to noise and artifacts, and elucidate the effects of signal processing and feature extraction on the accuracy of detection, classification, and estimation of PA and APS.

Contexts of emission of non-signature whistles in bottlenose dolphins (Tursiops truncatus) under human care

2020 ◽  
Vol 181 ◽  
pp. 104255
Author(s):  
Juliana Lopez-Marulanda ◽  
Heiko G. Rödel ◽  
Nikolaas Colpaert ◽  
Sander Vanderheul ◽  
Olivier Adam ◽  
...  
Keyword(s):  
Tursiops Truncatus ◽  
Bottlenose Dolphins ◽  
Signature Whistles ◽  
Human Care

scholarly journals Water cut/salt content forecasting in oil wells using a novel data-driven approach

2019 ◽  
Vol 74 ◽  
pp. 68
Author(s):  
Rouhollah Ahmadi ◽  
Jamal Shahrabi ◽  
Babak Aminshahidy
Keyword(s):  
Classification Tree ◽  
Salt Content ◽  
Oil Wells ◽  
Data Driven ◽  
Var Model ◽  
Support Vector ◽  
Water Production ◽  
Linear Discriminant ◽  
Data Driven Approach ◽  
Water Cut

Water cut is an important parameter in reservoir management and surveillance. Unlike traditional approaches, including numerical simulation and analytical techniques, which were developed for predicting water production in oil wells based on some assumptions and limitations, a new data-driven approach is proposed for forecasting water cut in two different types of oil wells in this article. First, a classification approach is presented for water cut prediction in sweet oil wells with discontinuous salt production patterns. Different classification algorithms including Support Vector Machine (SVM), Classification Tree (CT), Random Forest (RF), Multi-Layer Perceptron (MLP), Linear Discriminant Analysis (LDA) and Naïve Bayes (NB) are investigated in this regard. According to the results of a case study on a real Iranian sweet oil well, RF, CT, MLP and SVM can provide the best performance measures, respectively. Next, a Vector Autoregressive (VAR) model is proposed for forecasting water cut in salty oil wells with continuous water production during the life of the well. The proposed VAR model is verified using data of two real salty oil wells. The results confirm that the well-tuned proposed VAR model could provide reliable and acceptable results with very good accuracy in forecasting water production for the near future days.

scholarly journals Data Mining Techniques for Analysing Employment Data

2019 ◽  
Vol 9 (2) ◽  
pp. 555-556
Keyword(s):  
Process Model ◽  
Large Scale ◽  
Classification Tree ◽  
Predictive Modelling ◽  
Testing Procedure ◽  
Support Vector ◽  
Linear Discriminant ◽  
Rigorous Testing ◽  
Data Partitions ◽  
Binary Classifiers

This paper proposes a methodology that uses a large-scale employment dataset in order to explore which factors affect employment and how. The proposed methodology is a combination of predictive modelling, variable significance analysis, and VEC analysis. Modelling is based on logistic regression, linear discriminant analysis, neural network, classification tree, and support vector machine. Following the CRISP-DM standard process model, we train binary classifiers optimising their hyper-parameters and measure their performance by prediction accuracy, ROC analysis, and AUC. Using sensitivity analysis, we rank the variable significance in order to identify and measure factors of employment. Using VEC analysis, we further explore how values of those factors affect employment. Findings show that best performing models are neural networks and support vector machines with preference to the latter for quality of VEC. Experiments also suggest that education and age are primary contributors for correct classification with specific value distribution, discussed in the paper. All results were validated using a rigorous testing procedure that involves training, validation, and test data partitions and a combination of multiple runs along with three-fold cross-validation. This study addresses some gaps in previous research publications, which lack quantification of the conclusions made.

scholarly journals Probabilistic approaches for classifying highly variable anti-SARS-CoV-2 antibody responses

2020 ◽  
Author(s):  
Xaquin C Dopico ◽  
Leo Hanke ◽  
Daniel J. Sheward ◽  
Sandra Muschiol ◽  
Soo Aleman ◽  
...  
Keyword(s):  
Negative Control ◽  
Antibody Responses ◽  
Training Data ◽  
Support Vector ◽  
Correct Classification ◽  
Probabilistic Algorithms ◽  
Past Infection ◽  
Linear Discriminant

AbstractAntibody responses vary widely between individuals1, complicating the correct classification of low-titer measurements using conventional assay cut-offs. We found all participants in a clinically diverse cohort of SARS-CoV-2 PCR+ individuals (n=105) – and n=33 PCR+ hospital staff – to have detectable IgG specific for pre-fusion-stabilized spike (S) glycoprotein trimers, while 98% of persons had IgG specific for the receptor-binding domain (RBD). However, anti-viral IgG levels differed by several orders of magnitude between individuals and were associated with disease severity, with critically ill patients displaying the highest anti-viral antibody titers and strongest in vitro neutralizing responses. Parallel analysis of random healthy blood donors and pregnant women (n=1,000) of unknown serostatus, further demonstrated highly variable IgG titers amongst seroconverters, although these were generally lower than in hospitalized patients and included several measurements that scored between the classical 3 and 6SD assay cut-offs. Since the correct classification of seropositivity is critical for individual- and population-level metrics, we compared different probabilistic algorithms for their ability to assign likelihood of past infection. To do this, we used tandem anti-S and -RBD IgG responses from our PCR+ individuals (n=138) and a large cohort of historical negative controls (n=595) as training data, and generated an equal-weighted learner from the output of support vector machines and linear discriminant analysis. Applied to test samples, this approach provided a more quantitative way to interpret anti-viral titers over a large continuum, scrutinizing measurements overlapping the negative control background more closely and offering a probability-based diagnosis with potential clinical utility. Especially as most SARS-CoV-2 infections result in asymptomatic or mild disease, these platform-independent approaches improve individual and epidemiological estimates of seropositivity, critical for effective management of the pandemic and monitoring the response to vaccination.

Feature Reduction for Computationally Efficient Damage State Classification Using Binary Tree Support Vector Machines

2008 ◽  
Author(s):  
Clyde Coelho ◽  
Aditi Chattopadhyay
Keyword(s):  
Support Vector Machines ◽  
Binary Tree ◽  
Feature Reduction ◽  
Training Data ◽  
Support Vector ◽  
Computationally Efficient ◽  
Damage State ◽  
Data Set ◽  
Linear Discriminant ◽  
Vector Machines

This paper proposes a computationally efficient methodology for classifying damage in structural hotspots. Data collected from a sensor instrumented lug joint subjected to fatigue loading was preprocessed using a linear discriminant analysis (LDA) to extract features that are relevant for classification and reduce the dimensionality of the data. The data is then reduced in the feature space by analyzing the structure of the mapped clusters and removing the data points that do not affect the construction of interclass separating hyperplanes. The reduced data set is used to train a support vector machines (SVM) based classifier and the results of the classification problem are compared to those when the entire data set is used for training. To further improve the efficiency of the classification scheme, the SVM classifiers are arranged in a binary tree format to reduce the number of comparisons that are necessary. The experimental results show that the data reduction does not reduce the ability of the classifier to distinguish between classes while providing a nearly fourfold decrease in the amount of training data processed.

scholarly journals Identification and Characteristics of Signature Whistles in Wild Bottlenose Dolphins (Tursiops truncatus) from Namibia

PLoS ONE ◽  
2014 ◽  
Vol 9 (9) ◽  
pp. e106317 ◽  
Author(s):  
Hannah Joy Kriesell ◽  
Simon Harvey Elwen ◽  
Aurora Nastasi ◽  
Tess Gridley
Keyword(s):  
Tursiops Truncatus ◽  
Bottlenose Dolphins ◽  
Signature Whistles
Sign in / Sign up

Export Citation Format

Share Document