Identifying Microphone from Noisy Recordings by Using Representative Instance One Class-Classification Approach

Abstract

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

Fingerprinting Smartphones Based on Microphone Characteristics From Environment Affected Recordings

“…Audio signal characteristics, e.g., mean, standard deviation, dynamic range D, the crest-factor Q and autocorrelation time are analyzed in [24] within the scope of forensics applications. In [14], one-class classification is used based on noise collected from different locations, i.e., indoors or outdoors, inside a park or on a busy street. Characteristics extracted from FFT coefficients are used in [15] along with machine learning algorithms, i.e., Naive Bayes, multi class SVM, decision trees and KNN.…”

“…Type of sound Classifiers Devices Sound [8] 1kHz and 2kHz tone SVM, KNN, CNN 32 smartphones [9] 1kHz tone SVM, KNN, CNN 34 smartphones [10] 1kHz tone, pneumatic hammer, gunshot SVM, KNN, CNN 34 smartphones [11] 13 tones in the range of 100Hz-1300Hz maximum-likelihood classification 16 smartphones [12] 80 tones in the range of 100Hz-8kHz artificial neural networks 6 commercial microphones [13] ambient noise generated with a fan cooler inter-class cross correlation 8 commercial microphones [14] noise: indoor, park, street one-class classification 5 commercial microphones [15] music the smartphone loudspeaker is identified based on natural sounds, i.e., instrumental, song and human speech using distinct audio features, i.e., RMS (root-mean-square), ZCR (zero crossings), Low-Energy-Rate, Spectral Centroid, Spectral Entropy, etc. In [31], the Euclidean distance is used for the smartphones loudspeaker identification based on cosine tones between 14kHz and 21kHz with 100Hz increment.…”

Audio recording device identification based on deep learning

“…Sohaib Ikram et al [9] had a great idea about leakage signal, which is actually in the removed noise from speech enhancement, and we find the idea really inspiring. Huy Quan Vu et al [10] identified microphone from noisy recordings by using representative instance One Class-Classification approach, and proposed a representative instance classification framework to improve performance of OCC algorithms. Chang-Bae Moon et al [11] proposed an audio recorder identification method as one of digital forensic technologies, as well as a new feature reduction method, where Wiener filter was used to extract noise sounds of recorders and their features were extracted by MIRtoolbox.…”

Statistical Based Audio Forensic on Identical Microphones

“…They considered two features named MFCC and linear-cepstral coefficients. The most recent work was performed by Vu et al [17] in 2012. Vu et al [17] introduced novel approach called One-class classification (OCC) along with representative instance classification framework (RICF) for microphone forensics.…”

“…The most recent work was performed by Vu et al [17] in 2012. Vu et al [17] introduced novel approach called One-class classification (OCC) along with representative instance classification framework (RICF) for microphone forensics. The RICF is introduced to reduce the noisy signal such that it improves OCC performance.…”