Visualizing Audio for Anomaly Detection

Slides:



Advertisements
Similar presentations
Change-Point Detection Techniques for Piecewise Locally Stationary Time Series Michael Last National Institute of Statistical Sciences Talk for Midyear.
Advertisements

Does one size really fit all? Evaluating classifiers in Bag-of-Visual-Words classification Christian Hentschel, Harald Sack Hasso Plattner Institute.
Detecting Faces in Images: A Survey
Kaggle: Whale Challenge
Basic Spectrogram Lab 8. Spectrograms §Spectrograph: Produces visible patterns of acoustic energy called spectrograms §Spectrographic Analysis: l Acoustic.
Speaker Adaptation for Vowel Classification
Object Recognition by Parts Object recognition started with line segments. - Roberts recognized objects from line segments and junctions. - This led to.
Learning-Based Anomaly Detection in BGP Updates Jian Zhang Jennifer Rexford Joan Feigenbaum.
Robust Real-Time Object Detection Paul Viola & Michael Jones.
Object Recognition by Parts Object recognition started with line segments. - Roberts recognized objects from line segments and junctions. - This led to.
CIVS, Statistics Dept. UCLA Deformable Template as Active Basis Zhangzhang Si UCLA Department of Statistics Ying Nian Wu, Zhangzhang Si, Chuck.
Anomaly detection Problem motivation Machine Learning.
LE 460 L Acoustics and Experimental Phonetics L-13
Information Design and Visualization
SVCL Automatic detection of object based Region-of-Interest for image compression Sunhyoung Han.
Window-based models for generic object detection Mei-Chen Yeh 04/24/2012.
Lecture 29: Face Detection Revisited CS4670 / 5670: Computer Vision Noah Snavely.
1 PATTERN COMPARISON TECHNIQUES Test Pattern:Reference Pattern:
MUMT611: Music Information Acquisition, Preservation, and Retrieval Presentation on Timbre Similarity Alexandre Savard March 2006.
Structure Discovery of Pop Music Using HHMM E6820 Project Jessie Hsu 03/09/05.
A Comparative Study of Kernel Methods for Classification Applications Yan Liu Oct 21, 2003.
Applying Statistical Machine Learning to Retinal Electrophysiology Matt Boardman January, 2006 Faculty of Computer Science.
Interactive Learning of the Acoustic Properties of Objects by a Robot
Singer similarity / identification Francois Thibault MUMT 614B McGill University.
Judith C. Brown Journal of the Acoustical Society of America,1991 Jain-De,Lee.
Trust Me, I’m Partially Right: Incremental Visualization Lets Analysts Explore Large Datasets Faster Shengliang Dai.
Singer Similarity Doug Van Nort MUMT 611. Goal Determine Singer / Vocalist based on extracted features of audio signal Classify audio files based on singer.
Predicting Voice Elicited Emotions
Speaker Verification Using Adapted GMM Presented by CWJ 2000/8/16.
CS332 Visual Processing Department of Computer Science Wellesley College High-Level Vision Face Recognition I.
Phone-Level Pronunciation Scoring and Assessment for Interactive Language Learning Speech Communication, 2000 Authors: S. M. Witt, S. J. Young Presenter:
Zhiyao Duan, Changshui Zhang Department of Automation Tsinghua University, China Music, Mind and Cognition workshop.
Melinda Feldmann Combination Tones. What is a Combination Tone? Combination Tone In musical acoustics, faint tone produced in the inner ear by two simultaneously.
 Mentor : Prof. Amitabha Mukerjee Learning to Detect Salient Objects Team Members - Avinash Koyya Diwakar Chauhan.
A Tutorial on Speaker Verification First A. Author, Second B. Author, and Third C. Author.
CS 445/656 Computer & New Media
Object Recognition by Parts
Electronic Visualization Laboratory University of Illinois at Chicago
Large-Scale Music Audio Analyses Using High Performance Computing Technologies: Creating New Tools, Posing New Questions J.
PATTERN COMPARISON TECHNIQUES
Landmark-Based Speech Recognition: Spectrogram Reading, Support Vector Machines, Dynamic Bayesian Networks, and Phonology Mark Hasegawa-Johnson
Detecting Semantic Concepts In Consumer Videos Using Audio Junwei Liang, Qin Jin, Xixi He, Gang Yang, Jieping Xu, Xirong Li Multimedia Computing Lab,
The Q Pipeline search for gravitational-wave bursts with LIGO
A Forest of Sensors: Using adaptive tracking to classify and monitor activities in a site Eric Grimson AI Lab, Massachusetts Institute of Technology
Presentation on Artificial Neural Network Based Pathological Voice Classification Using MFCC Features Presenter: Subash Chandra Pakhrin 072MSI616 MSC in.
Potter’s Wheel: An Interactive Data Cleaning System
Filtering Geophysical Data: Be careful!
Multimedia: making it Work
Context-based vision system for place and object recognition
Object Recognition by Parts
A Tutorial on HOG Human Detection
Outlier Discovery/Anomaly Detection
Face detection using Random projections
Pitch Detection from Waveform and Spectrogram
Object Recognition by Parts
Object Recognition by Parts
Information Design and Visualization
Brief Review of Recognition + Context
Working with Multimedia
Anomaly Detection in Crowded Scenes
3-D Model Tips & Preparing and Practicing the Team Presentation
Audio and Speech Computers & New Media.
AUDIO SURVEILLANCE SYSTEMS: SUSPICIOUS SOUND RECOGNITION
Machine Learning in Practice Lecture 27
ECE 791 Project Proposal Project Title: Developing and Evaluating a Tool for Converting MP3 Audio Files to Staff Music Project Team: Salvatore DeVito.
Jia-Bin Huang Virginia Tech
Object Recognition by Parts
Midshipman grunt trains.
Object Recognition with Interest Operators
Speech Prosody Conversion using Sequence Generative Adversarial Nets
Presentation transcript:

Visualizing Audio for Anomaly Detection Mark Hasegawa-Johnson Camille Goudeseune Hank Kaczmarski Thomas Huang University of Illinois at Urbana-Champaign

Research Goal: Guide audio analysts to anomalies Large dataset: audio Anomalies Cheap to record, expensive to play GUI: listen 10000x faster Robots are poor listeners, but good servants

Anomalies shoot down poor theories

Feature : audio interval  numbers Visualization : numbers  rendering Audio-based features (spectrogram) Model-based features (Hnull, Hthreatening)

Audio-Based Features: Transformed acoustic data Pitch, Formants Anomaly Salience Score Waveform A = f(t) Spectrogram A = f(t, Hz) Correlogram A = f(t, Hz, fundamental period) Rate-scale representation A = f(t, Hz, bandwidth, ΔHz) Wavelets Multiscale

Model-Based Features Log-likelihood features Display how well a hypothesis fits Let analyst intuit a threshold of fitness Defined by Mixture Gaussians, trained by: EM Parzen Windows One-class SVMs

Model-Based Features Log-likelihood features  Log-likelihood ratios (LLR) Because Mixture Gaussian misclassifies outliers as anomalies

Too many features! Evaluate each feature with Kullback-Leibler Divergence Combine features with AdaBoost + SVM + HMM

Two Interactive Testbeds Vary features Vary anomalies Vary background audio Vary how model is trained Vary mapping from features to HSV Anomalousness “bubbles up”

Multi-day audio timeline

1000  μphone = The Milliphone

Human Subject Protocols Tutorial Training with immediate feedback Measure how fast subjects find x% of anomalies

Influence on FODAVA Guide, don’t replace, human analysts Guide them with zoomable features Features from transformed data (audio-based) Features from fitting hypotheses (model-based)

Developing FODAVA Make “big” audio accessible Audio is hard, but its concepts generalize Fast interactive exploration of time series, long (timeline) or wide (milliphone)