scholarly journals Evaluation of speaker de-identification based on voice gender and age conversion

2018 ◽  
Vol 69 (2) ◽  
pp. 138-147 ◽  
Author(s):  
Jiří Přibil ◽  
Anna Přibilová ◽  
Jindřich Matoušek
Keyword(s):  
Gaussian Mixture Models ◽  
Gaussian Mixture ◽  
Scale Transformation ◽  
English Speakers ◽  
Detection Accuracy ◽  
Frequency Scale ◽  
Open Set ◽  
Gender And Age ◽  
Prosody Modification ◽  
Apparent Age

AbstractTwo basic tasks are covered in this paper. The first one consists in the design and practical testing of a new method for voice de-identification that changes the apparent age and/or gender of a speaker by multi-segmental frequency scale transformation combined with prosody modification. The second task is aimed at verification of applicability of a classifier based on Gaussian mixture models (GMM) to detect the original Czech and Slovak speakers after applied voice deidentification. The performed experiments confirm functionality of the developed gender and age conversion for all selected types of de-identification which can be objectively evaluated by the GMM-based open-set classifier. The original speaker detection accuracy was compared also for sentences uttered by German and English speakers showing language independence of the proposed method.

scholarly journals Deep-Learning-Based Road Crack Detection Frameworks for Dashcam-captured Images under Different Illumination Conditions

2021 ◽  
Author(s):  
Da-Ren Chen ◽  
Wei-Min Chiu
Keyword(s):  
Object Detection ◽  
Large Scale ◽  
Crack Detection ◽  
State Of The Art ◽  
Gaussian Mixture Models ◽  
Gaussian Mixture ◽  
Machine Learning Techniques ◽  
Detection Accuracy ◽  
The Road ◽  
Art Object

Abstract Machine learning techniques have been used to increase detection accuracy of cracks in road surfaces. Most studies failed to consider variable illumination conditions on the target of interest (ToI), and only focus on detecting the presence or absence of road cracks. This paper proposes a new road crack detection method, IlumiCrack, which integrates Gaussian mixture models (GMM) and object detection CNN models. This work provides the following contributions: 1) For the first time, a large-scale road crack image dataset with a range of illumination conditions (e.g., day and night) is prepared using a dashcam. 2) Based on GMM, experimental evaluations on 2 to 4 levels of brightness are conducted for optimal classification. 3) the IlumiCrack framework is used to integrate state-of-the-art object detecting methods with CNN to classify the road crack images into eight types with high accuracy. Experimental results show that IlumiCrack outperforms the state-of-the-art R-CNN object detection frameworks.

Autonomous Anomaly Detection System for Crime Monitoring

2019 ◽  
Vol 16 (8) ◽  
pp. 3410-3418
Author(s):  
Muhammed Shuaau ◽  
Ka Fei Thang
Keyword(s):  
Anomaly Detection ◽  
Trajectory Analysis ◽  
Detection System ◽  
False Positive Rate ◽  
Gaussian Mixture Models ◽  
Gaussian Mixture ◽  
Asia Pacific ◽  
Detection Accuracy ◽  
Positive Rate ◽  
Reception Area

Autonomous anomaly detection has attracted significant amount of attention in the past decade due to increased security concerns all around the world. The volume of data reported by surveillance cameras has outrun human capacity and there exists a greater need for anomaly detection systems for crime monitoring. This project proposes a solution to this problem in a reception area context by using trajectory analysis. Trajectory extraction is proposed by using Gaussian Mixture Models and Kalman Filter for data association. Then trajectory analysis is performed on extracted trajectories to detect four different anomalies which are entering staff area, running, loitering and squatting down. The proposed anomaly detection method is tested on datasets recorded at Asia Pacific University’s reception area. The proposed algorithms were able to achieve a detection accuracy of 89% and a false positive rate of 4.52%. The results presented show the effectiveness of the proposed method.

scholarly journals Autonomous Abnormal Behaviour Detection Using Trajectory Analysis

2019 ◽  
Vol 9 (4) ◽  
pp. 2403
Author(s):  
Muhammed Shuaau ◽  
Ka Fei Thang ◽  
Nai Shyan Lai
Keyword(s):  
Anomaly Detection ◽  
Trajectory Analysis ◽  
False Positive Rate ◽  
Gaussian Mixture Models ◽  
Gaussian Mixture ◽  
Asia Pacific ◽  
Detection Accuracy ◽  
Abnormal Behaviour ◽  
Positive Rate ◽  
Reception Area

<span lang="EN-GB">Abnormal behaviour detection has attracted signification amount of attention in the past decade due to increased security concerns around the world. The amount of data from surveillance cameras have exceeded human capacity and there is a greater need for anomaly detection systems for crime monitoring. This paper proposes a solution to this problem in a reception area context by using trajectory extraction through Gaussian Mixture Models and Kalman Filter for data association. Here, trajectory analysis was performed on extracted trajectories to detect four different anomalies such as entering staff area, running, loitering and squatting down. The developed anomaly detection algorithms were tested on videos captured at Asia Pacific University’s reception area. These algorithms were able to achieve a promising detection accuracy of 89% and a false positive rate of 4.52%.</span>

An Optimization Algorithm of Foreground Objects Extraction

2011 ◽  
Vol 403-408 ◽  
pp. 169-176
Author(s):  
Xia Yi Zhang ◽  
Zhi Peng Li ◽  
Fu Qiang Liu ◽  
Zhen Jia ◽  
Jian Wei Zhao
Keyword(s):  
Background Subtraction ◽  
Gaussian Mixture Models ◽  
Computational Cost ◽  
Gaussian Mixture ◽  
Object Extraction ◽  
Detection Accuracy ◽  
Foreground Object ◽  
Image Matting ◽  
New Approach ◽  
Novel Approach

In this paper, we propose a novel algorithm for coarse-to-fine foreground objects extraction. There are two general approaches for foreground objects extraction: background subtraction and image matting. Our new approach can not only improve detection accuracy compared with general background subtraction approaches, but also reduce computation burden compared with general image matting approaches. Firstly, we present a novel method called Motion-mask Gaussian Mixture Models (Motion-mask GMMs) to extract coarse foreground regions. This new approach can classify foreground and background pixels more accurately, especially when there are long-time stopping objects in the scene. Secondly, with the coarse foreground regions, we propose a novel approach to make foreground object extraction more accurate based on effective fusion of image registration and image matting. This new method overcomes the template drift problem during template updating and also reduces the expensive computational cost of image matting. Our proposed approach is tested with kinds of video sequences in indoor and outdoor environments. Experimental results demonstrate the accuracy and efficiency of our proposed approach for foreground object extraction.

scholarly journals Measure of Similarity between GMMs by Embedding of the Parameter Space That Preserves KL Divergence

Mathematics ◽  
2021 ◽  
Vol 9 (9) ◽  
pp. 957
Author(s):  
Branislav Popović ◽  
Lenka Cepova ◽  
Robert Cep ◽  
Marko Janev ◽  
Lidija Krstanović
Keyword(s):  
Computational Complexity ◽  
Parameter Space ◽  
Recognition Accuracy ◽  
Gaussian Mixture Models ◽  
Gaussian Mixture ◽  
Dimensional Manifold ◽  
Dimensional Parameter ◽  
Trade Off ◽  
Measure Of Similarity ◽  
Lower Dimensional

In this work, we deliver a novel measure of similarity between Gaussian mixture models (GMMs) by neighborhood preserving embedding (NPE) of the parameter space, that projects components of GMMs, which by our assumption lie close to lower dimensional manifold. By doing so, we obtain a transformation from the original high-dimensional parameter space, into a much lower-dimensional resulting parameter space. Therefore, resolving the distance between two GMMs is reduced to (taking the account of the corresponding weights) calculating the distance between sets of lower-dimensional Euclidean vectors. Much better trade-off between the recognition accuracy and the computational complexity is achieved in comparison to measures utilizing distances between Gaussian components evaluated in the original parameter space. The proposed measure is much more efficient in machine learning tasks that operate on large data sets, as in such tasks, the required number of overall Gaussian components is always large. Artificial, as well as real-world experiments are conducted, showing much better trade-off between recognition accuracy and computational complexity of the proposed measure, in comparison to all baseline measures of similarity between GMMs tested in this paper.

scholarly journals Detection of Unknown DDoS Attacks with Deep Learning and Gaussian Mixture Model

2021 ◽  
Vol 11 (11) ◽  
pp. 5213
Author(s):  
Chin-Shiuh Shieh ◽  
Wan-Wei Lin ◽  
Thanh-Tuan Nguyen ◽  
Chi-Hong Chen ◽  
Mong-Fong Horng ◽  
...  
Keyword(s):  
Deep Learning ◽  
Gaussian Mixture Model ◽  
Mixture Model ◽  
Denial Of Service ◽  
Gaussian Mixture ◽  
Mitigation Measures ◽  
Distribution Model ◽  
Ddos Attacks ◽  
Open Set ◽  
The Impact

DDoS (Distributed Denial of Service) attacks have become a pressing threat to the security and integrity of computer networks and information systems, which are indispensable infrastructures of modern times. The detection of DDoS attacks is a challenging issue before any mitigation measures can be taken. ML/DL (Machine Learning/Deep Learning) has been applied to the detection of DDoS attacks with satisfactory achievement. However, full-scale success is still beyond reach due to an inherent problem with ML/DL-based systems—the so-called Open Set Recognition (OSR) problem. This is a problem where an ML/DL-based system fails to deal with new instances not drawn from the distribution model of the training data. This problem is particularly profound in detecting DDoS attacks since DDoS attacks’ technology keeps evolving and has changing traffic characteristics. This study investigates the impact of the OSR problem on the detection of DDoS attacks. In response to this problem, we propose a new DDoS detection framework featuring Bi-Directional Long Short-Term Memory (BI-LSTM), a Gaussian Mixture Model (GMM), and incremental learning. Unknown traffic captured by the GMM are subject to discrimination and labeling by traffic engineers, and then fed back to the framework as additional training samples. Using the data sets CIC-IDS2017 and CIC-DDoS2019 for training, testing, and evaluation, experiment results show that the proposed BI-LSTM-GMM can achieve recall, precision, and accuracy up to 94%. Experiments reveal that the proposed framework can be a promising solution to the detection of unknown DDoS attacks.

scholarly journals A Unified Formulation of k-Means, Fuzzy c-Means and Gaussian Mixture Model by the Kolmogorov–Nagumo Average

Entropy ◽  
2021 ◽  
Vol 23 (5) ◽  
pp. 518
Author(s):  
Osamu Komori ◽  
Shinto Eguchi
Keyword(s):  
Pareto Distribution ◽  
Statistical Data ◽  
Learning Algorithm ◽  
Survival Function ◽  
Gaussian Mixture Models ◽  
Gaussian Mixture ◽  
Simulation Studies ◽  
Probabilistic Framework ◽  
Underlying Distribution ◽  
Fuzzy C Means

Clustering is a major unsupervised learning algorithm and is widely applied in data mining and statistical data analyses. Typical examples include k-means, fuzzy c-means, and Gaussian mixture models, which are categorized into hard, soft, and model-based clusterings, respectively. We propose a new clustering, called Pareto clustering, based on the Kolmogorov–Nagumo average, which is defined by a survival function of the Pareto distribution. The proposed algorithm incorporates all the aforementioned clusterings plus maximum-entropy clustering. We introduce a probabilistic framework for the proposed method, in which the underlying distribution to give consistency is discussed. We build the minorize-maximization algorithm to estimate the parameters in Pareto clustering. We compare the performance with existing methods in simulation studies and in benchmark dataset analyses to demonstrate its highly practical utilities.

Sign in / Sign up

Export Citation Format

Share Document