Microphone identification using one-class classification approach

Rapid growth of technical developments has created huge challenges for microphone forensics - a subcategory of audio forensic science, because of the availability of numerous digital recording devices and massive amount of recording data. Demand for fast and efficient methods to assure integrity and authenticity of information is becoming more and more important in criminal investigation nowadays. Machine learning has emerged as an important technique to support audio analysis processes of microphone forensic practitioners. However, its application to real life situations using supervised learning is still facing great challenges due to expensiveness in collecting data and updating system. In this paper, we introduce a new machine learning approach which is called One-class Classification (OCC) to be applied to microphone forensics; we demonstrate its capability on a corpus of audio samples collected from several microphones. Research results and analysis indicate that OCC has the potential to benefit microphone forensic practitioners in developing new tools and techniques for effective and efficient analysis.

Identifying microphone from noisy recordings by using representative instance one class-classification approach

Rapid growth of technical developments has created huge challenges for microphone forensics - a sub-category of audio forensic science, because of the availability of numerous digital recording devices and massive amount of recording data. Demand for fast and efficient methods to assure integrity and authenticity of information is becoming more and more important in criminal investigation nowadays. Machine learning has emerged as an important technique to support audio analysis processes of microphone forensic practitioners. However, its application to real life situations using supervised learning is still facing great challenges due to expensiveness in collecting data and updating system. In this paper, we introduce a new machine learning approach which is called One-class Classification (OCC) to be applied to microphone forensics; we demonstrate its capability on a corpus of audio samples collected from several microphones. In addition, we propose a representative instance classification framework (RICF) that can effectively improve performance of OCC algorithms for recording signal with noise. Experiment results and analysis indicate that OCC has the potential to benefit microphone forensic practitioners in developing new tools and techniques for effective and efficient analysis.

Multi-label approach for human-face classification

Single-label classification models have been widely used for human-face classification. In this paper, we present a multi-label classification approach for human-face classification. Multi-label classification is more appropriate in the real world because a human-face can be associated with multiple labels. Demographic information can be derived and utilized along with facial expression in the field of face classification to assist with multi label classification. Gabor filters; Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) methods, are used to extract and project representative demographic information from facial images. For evaluation, five classification algorithms were used. We evaluate the proposed approach by performing experiments on Yale face images database. Results show the effectiveness of multi-label classification algorithms.

A case study on classification reliability

The reliability of an induced classifier can be affected by several factors including the data oriented factors and the algorithm oriented factors. In some cases, the reliability could also be affected by knowledge oriented factors. In this paper, we analyze three special cases to examine the reliability of the discovered knowledge. Our case study results show that (1) in the cases of mining from low quality data, rough classification approach is more reliable than exact approach which in general tolerate to low quality data; (2) Without sufficient large size of the data, the reliability of the discovered knowledge will be decreased accordingly; (3) The reliability of point learning approach could easily be misled by noisy data. It will in most cases generate an unreliable interval and thus affect the reliability of the discovered knowledge. It is also reveals that the inexact field is a good learning strategy that could model the potentials and to improve the discovery reliability.

Email categorization using (2+1)-tier classification algorithms

In this paper we have proposed a spam filtering technique using (2+1)-tier classification approach. The main focus of this paper is to reduce the false positive (FP) rate which is considered as an important research issue in spam filtering. In our approach, firstly the email message will classify using first two tier classifiers and the outputs will appear to the analyzer. The analyzer will check the labeling of the output emails and send to the corresponding mailboxes based on labeling, for the case of identical prediction. If there are any misclassifications occurred by first two tier classifiers then tier-3 classifier will invoked by the analyzer and the tier-3 will take final decision. This technique reduced the analyzing complexity of our previous work. It has also been shown that the proposed technique gives better performance in terms of reducing false positive as well as better accuracy.

The impact of sample size and data quality to classification reliability

The reliability of an induced classifier can be affected by several factors including the data oriented factors and the algorithm oriented factors [3]. In some cases, the reliability could also be affected by knowledge oriented factors. In this chapter, we analyze three special cases to examine the reliability of the discovered knowledge. Our case study results show that (1) in the cases of mining from low quality data, rough classification approach is more reliable than exact approach which in general tolerate to low quality data; (2) Without sufficient large size of the data, the reliability of the discovered knowledge will be decreased accordingly; (3) The reliability of point learning approach could easily be misled by noisy data. It will in most cases generate an unreliable interval and thus affect the reliability of the discovered knowledge. It is also reveals that the inexact field is a good learning strategy that could model the potentials and to improve the discovery reliability.

Automated classification of human daily activities in ambulatory environment

This paper presents a human daily activity classification approach based on the sensory data collected from a single tri-axial accelerometer worn on waist belt. The classification algorithm was realized to distinguish 6 different activities including standing, jumping, sitting-down, walking, running and falling through three major steps: wavelet transformation, Principle Component Analysis (PCA)-based dimensionality reduction and followed by implementing a radial basis function (RBF) kernel Support Vector Machine (SVM) classifier. Two trials were conducted to evaluate different aspects of the classification scheme. In the first trial, the classifier was trained and evaluated by using a dataset of 420 samples collected from seven subjects by using a k-fold cross-validation method. The parameters σ and c of the RBF kernel were optimized through automatic searching in terms of yielding the highest recognition accuracy and robustness. In the second trial, the generation capability of the classifier was also validated by using the dataset collected from six new subjects. The average classification rates of 95% and 93% are obtained in trials 1 and 2, respectively. The results in trial 2 show the system is also good at classifying activity signals of new subjects. It can be concluded that the collective effects of the usage of single accelerometer sensing, the setting of the accelerometer placement and efficient classifier would make this wearable sensing system more realistic and more comfortable to be implemented for long-term human activity monitoring and classification in ambulatory environment, therefore, more acceptable by users.

Network traffic classification using correlation information

Traffic classification has wide applications in network management, from security monitoring to quality of service measurements. Recent research tends to apply machine learning techniques to flow statistical feature based classification methods. The nearest neighbor (NN)-based method has exhibited superior classification performance. It also has several important advantages, such as no requirements of training procedure, no risk of overfitting of parameters, and naturally being able to handle a huge number of classes. However, the performance of NN classifier can be severely affected if the size of training data is small. In this paper, we propose a novel nonparametric approach for traffic classification, which can improve the classification performance effectively by incorporating correlated information into the classification process. We analyze the new classification approach and its performance benefit from both theoretical and empirical perspectives. A large number of experiments are carried out on two real-world traffic data sets to validate the proposed approach. The results show the traffic classification performance can be improved significantly even under the extreme difficult circumstance of very few training samples.

Habitat classification of temperate marine macroalgal communities using bathymetric LiDAR

Here, we evaluated the potential of using bathymetric Light Detection and Ranging (LiDAR) to characterise shallow water (<30 m) benthic habitats of high energy subtidal coastal environments. Habitat classification, quantifying benthic substrata and macroalgal communities, was achieved in this study with the application of LiDAR and underwater video groundtruth data using automated classification techniques. Bathymetry and reflectance datasets were used to produce secondary terrain derivative surfaces (e.g., rugosity, aspect) that were assumed to influence benthic patterns observed. An automated decision tree classification approach using the Quick Unbiased Efficient Statistical Tree (QUEST) was applied to produce substrata, biological and canopy structure habitat maps of the study area. Error assessment indicated that habitat maps produced were primarily accurate (>70%), with varying results for the classification of individual habitat classes; for instance, producer accuracy for mixed brown algae and sediment substrata, was 74% and 93%, respectively. LiDAR was also successful for differentiating canopy structure of macroalgae communities (i.e., canopy structure classification), such as canopy forming kelp versus erect fine branching algae. In conclusion, habitat characterisation using bathymetric LiDAR provides a unique potential to collect baseline information about biological assemblages and, hence, potential reef connectivity over large areas beyond the range of direct observation. This research contributes a new perspective for assessing the structure of subtidal coastal ecosystems, providing a novel tool for the research and management of such highly dynamic marine environments. © 2014 by the authors; licensee MDPI, Basel, Switzerland.

Hydro-acoustic remote sensing of benthic biological communities on the shallow South East Australian continental shelf

Information regarding the composition and extent of benthic habitats on the South East Australian continental shelf is limited. In this habitat mapping study, multibeam echosounder (MBES) data are integrated with precisely geo-referenced video ground-truth data to quantify benthic biotic communities at Cape Nelson, Victoria, Australia. Using an automated decision tree classification approach, 5 representative biotic groups defined from video analysis were related to hydro-acoustically derived variables in the Cape Nelson survey area. Using a combination of multibeam bathymetry, backscatter and derivative products produced highest overall accuracy (87%) and kappa statistic (0.83). This study demonstrates that decision tree classifiers are capable of integrating variable data types for mapping distributions of benthic biological assemblages, which are important in maintaining biodiversity and other system services in the marine environment.

Corner inlet and Nooramunga habitat mapping project

Deakin University and the University of Tasmania were commissioned by Parks Victoria (PV) to create two updated habitat maps for areas within the Corner Inlet and Nooramunga Marine and Coastal Park and Ramsar area. The team obtained a ground-truth data set using in situ video and still photographs. This dataset was used to develop and assess predictive models of benthic marine habitat distributions incorporating data from both ALOS (Advanced Land Observation Satellite) imagery atmospherically corrected by CSIRO and LiDAR (Light Detection and Ranging) bathymetry. This report describes the results of the mapping effort as well as the methodology used to produce these habitat maps.

Overall accuracies of habitat classifications were good, returning overall accuracies >73 % and kappa values > 0.62 for both study localities. Habitats predicted with highest accuracies included Zosteraceae in Nooramunga (91 %), reef in Corner Inlet (80 %), and bare sediment (no-visible macrobiota/no-visible seagrass classes; both > 76 %). The majority of classification errors were due to the misclassification of areas of sparse seagrass as bare sediment. For the Corner Inlet study locality the no-visible macrobiota (10,698 ha), Posidonia (4,608 ha) and Zosteraceae (4,229 ha) habitat classes covered the most area. In Nooramunga no-visible seagrass (5,538 ha), Zosteraceae (4,060 ha) and wet saltmarsh (1,562 ha) habitat classes were most dominant.

In addition to the commissioned work preliminary change detection analyses were undertaken as part of this project. These analyses indicated shifts in habitat extents in both study localities since the late 1990s/2000. In particular, a post-classification analysis highlighted that there were considerable increases in seagrass habitat (primarily Zosteraceae) throughout the littoral zones and river/creek mouths of both study localities. Further, the numerous channel systems remained stable and were free of seagrass at both times. A substantial net loss of Posidonia in the Corner Inlet locality is likely but requires further investigation due to potential misclassifications between habitats in both the 1998 map (Roob et al. 1998) and the current mapping. While the unsupervised Independent Components Analysis (ICA) change detection technique indicated some changes in habitat extent and distribution, considerable areas of habitat change observed in the post-classification approach are questionable, and may reflect misclassifications rather than real change. A particular example of this is an apparent large decrease in Zosteraceae and increase in Posidonia being related to the classification of Posidonia beds as Zosteraceae in the 1998 mapping. Despite this, we believe that changes indicated by both the ICA and post-classification approaches have a high likelihood of being ‘actual’ change. A pattern of gains and losses of Zosteraceae in the region north of Stockyard channel is an example of this. Further analyses and refinements of approaches in change detection analyses such as would improve confidence in the location and extent of habitat changes over this time period.

This work has been successful in providing new baseline maps using a repeatable method meaning that any future changes in intertidal and shallow water marine habitats may be assessed in a consistent way with quantitative error assessments. In wider use, these maps should also allow improved conservation planning, advance fisheries and catchment management, and progress infrastructure planning to limit impacts on the Inlet environment.

Investigating machine learning techniques for detection of depression using structural MRI volumetric features

Structural MRI offers anatomical details and high sensitivity to pathological changes. It can demonstrate certain patterns of brain changes present at a structural level. Research to date has shown that volumetric analysis of brain regions has importance in depression detection. However, such analysis has had very minimal use in depression detection studies at individual level. Optimally combining various brain volumetric features/attributes, and summarizing the data into a distinctive set of variables remain difficult. This study investigates machine learning algorithms that automatically identify relevant data attributes for depression detection. Different machine learning techniques are studied for depression classification based on attributes extracted from structural MRI (sMRI) data. The attributes include volume calculated from whole brain, white matter, grey matter and hippocampus. Attributes subset selection is performed aiming to remove redundant attributes using three filtering methods and one hybrid method, in combination with ranker search algorithms. The highest average classification accuracy, obtained by using a combination of both SVM-EM and IG-Random Tree algorithms, is 85.23%. The classification approach implemented in this study can achieve higher accuracy than most reported studies using sMRI data, specifically for detection of depression.

Yaringa and French Island Marine National Park habitat mapping

Deakin University and the Department of Primary Industries were commissioned by ParksVictoria (PV) to create two updated habitat maps for Yaringa and French Island MarineNational Parks. The team obtained a ground-truth data set using in situ video and still photographs. This dataset was used to develop and assess predictive models of benthic marine habitat distributions incorporating data from World-View-2 imagery atmospherically corrected by CSIRO and LiDAR (Light Detection and Ranging) bathymetry. In addition, the team applied an unsupervised classification approach to an aerial photograph to assess the differences between the two remote sensors. This report describes the results of the mapping as well as the methodology used to produce these habitat maps.This study has provided mapping of intertidal and subtidal habitats of Yaringa and FrenchIsland MNPs at a 2 m resolution with fair to good accuracies (Kappa 0.40-0.75). These were combined with mangrove and saltmarsh habitats recently mapped by Boon et al. (2011) to provide compete-coverage habitat maps of Yaringa and French Island MNPs.The mapping showed that Yaringa MNP was dominated by mangroves, wet saltmarsh and dense Zostereaceae, covering 33%, 29% and 19%, respectively. Similarly, intertidalvegetation and subtidal vegetation (dominated by Zosteraceae) covered 26% and 25% ofFrench Island MNP. However, as a result of turbidity and missing satellite imagery 27% ofFrench Island MNP remains unmapped.The coupling of WV-2 and LiDAR reduced potential artefacts (e.g. sun glint causing whiteand black pixels known as the “salt and pepper effect”). The satellite classification appeared to provide better results than the aerial photography classification. However, since there is a two-year difference between the capture of the aerial photography and the collection of the ground-truth data this comparison is potentially temporally confounded. It must also be noted that there are differences in costs of the data,the spatial resolution between the two datasets (i.e. WV-2 = 2 m and the Aerial = 0.5 m) and the amount spectral information contained in the data (i.e. WV-2 = 8 bands and the aerial = 4 bands), which may ultimately determine its utility for a particular project.The spatial assessment using FRAGSTATs of habitat patches within Yaringa MNP provides a viable and cost effect way to assess habitat condition (i.e. shape, size and arrangement).This spatial assessment determined that dense Zosteraceae and NVSG habitat classeswere generally larger in patch size and continuity than the medium/sparse Zosteraceaehabitat. The application spatial techniques to time-series mapping may provide a way toremotely monitor the change in the spatial characteristics of marine habitats.This work was successful in providing new baseline habitat maps using a repeatable method meaning that any future changes in intertidal and shallow water marine habitats may be assessed in a consistent way with quantitative error assessments. In wider use, these maps should also allow improved conservation planning, fisheries and catchment management, and contribute toward infrastructure planning to limit impacts on Western Port.