Robust Speaker Identification Algorithms and Results in Noisy Environments

Ayhan, Bulent; Kwan, Chiman

doi:10.1007/978-3-319-92537-0_51

Cited by 6 publications

(7 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This problem is particularly expressed in very noisy environments similar to those in our study. Ayhan et al showed that by a combination of mask estimation, gammatone features with bounded marginalization dealing with unreliable features with a classic Gaussian mixture model may lead to an improvement in distinguishing the lead signal [31].…”

Section: Analysis Of Industrial Machinery Data For Predictive Maintenancementioning

confidence: 99%

Acoustic Anomaly Detection of Mechanical Failures in Noisy Real-Life Factory Environments

2021

View full text Add to dashboard Cite

Anomaly detection without employing dedicated sensors for each industrial machine is recognized as one of the essential techniques for preventive maintenance and is especially important for factories with low automatization levels, a number of which remain much larger than autonomous manufacturing lines. We have based our research on the hypothesis that real-life sound data from working industrial machines can be used for machine diagnostics. However, the sound data can be contaminated and drowned out by typical factory environmental sound, making the application of sound data-based anomaly detection an overly complicated process and, thus, the main problem we are solving with our approach. In this paper, we present a noise-tolerant deep learning-based methodology for real-life sound-data-based anomaly detection within real-world industrial machinery sound data. The main element of the proposed methodology is a generative adversarial network (GAN) used for the reconstruction of sound signal reconstruction and the detection of anomalies. The experimental results obtained in the Malfunctioning Industrial Machine Investigation and Inspection (MIMII) show the superiority of the proposed methodology over baseline approaches based on the One-Class Support Vector Machine (OC-SVM) and the Autoencoder–Decoder neural network. The proposed schematics using the unscented Kalman Filter (UKF) and the mean square error (MSE) loss function with the L2 regularization term showed an improvement of the Area Under Curve (AUC) for the noisy pump data of the pump.

show abstract

Section: Analysis Of Industrial Machinery Data For Predictive Maintenancementioning

confidence: 99%

Acoustic Anomaly Detection of Mechanical Failures in Noisy Real-Life Factory Environments

2021

View full text Add to dashboard Cite

show abstract

“…The results of their method of implementation and testing were increased up to 28% accuracy at signal to noise ratio (SNR) 5 dB. Ayhan and Kwan [21] developed a vigorous speaker identification scheme under noisy conditions which implicates "mask estimation, gammatone features with bounded marginalization to deal with unreliable features, and Gaussian mixture model (GMM) for speaker identification". Evaluation and assessments were performed to determine the speaker identification performance of the proposed algorithm, and results showed that it substantially outperforms the conventional method MFCC with Cepstral Mean Normalization (MFCC-CMN) at low signal-to-noise conditions.…”

Section: Introduction and Literature Reviewmentioning

confidence: 99%

“…Some scholars favored examining the use of CASA modules in noisy speech, in conjunction with one of the above-mentioned acoustic features, and results showed substantial improvement in identification performance in some cases. Moreover, many studies used the conventional classifiers, such as SVMs [13], GMMs [18], [21] and HMMs [23][24][25] , while many recent work explored the DNN-based classifiers [16], [26].…”

Section: Introduction and Literature Reviewmentioning

confidence: 99%

CASA-based speaker identification using cascaded GMM-CNN classifier in noisy and emotional talking conditions

Nassif

Shahin

Hamsa

et al. 2021

Applied Soft Computing

View full text Add to dashboard Cite

“…(2) The SST feature has better identification ability for complex modulated signals in comparison to the former methods, whereas its effect fluctuates when used to process simple modulated signals. In addition, some recognition methods in like acoustic and speech identification, like Gaussian Mixed Model (GMM) may also be migrated to SEI signal recognition [8,9]. However, there still exist some problems: GMM is based on a principle that the datasets should be large enough, unfortunately, it is not easy to realize in SEI, because radar signal acquisition needs higher experimental conditions, such as radar, microwave anechoic chamber, etc.…”

Section: Introductionmentioning

confidence: 99%

A Novel Data-Driven Specific Emitter Identification Feature Based on Machine Cognition

2020

View full text Add to dashboard Cite

Machine learning becomes increasingly promising in specific emitter identification (SEI), particularly in feature extraction and target recognition. Traditional features, such as radio frequency (RF), pulse amplitude (PA), power spectral density (PSD), and etc., usually show limited recognition effects when only a slight difference exists in radar signals. Numerous two-dimensional features on transform domain, like various time-frequency representation and ambiguity function are used to augment information abundance, whereas the unacceptable computational burden usually emerges. To solve this problem, some artfully handcrafted features in transformed domain are proposed, like representative slice of ambiguity function (AF-RS) and compressed sensing mask (CS-MASK), to extract representative information that contributes to machine recognition task. However, most handcrafted features only utilizing neural network as a classifier, few of them focus on mining deep informative features from the perspective of machine cognition. Such feature extraction that is based on human cognition instead of machine cognition may probably miss some seemingly nominal texture information which actually contributes greatly to recognition, or collect too much redundant information. In this paper, a novel data-driven feature extraction is proposed based on machine cognition (MC-Feature) resort to saliency detection. Saliency detection exhibits positive contributions and suppresses irrelevant contributions in a transform domain with the help of a saliency map calculated from the accumulated gradients of each neuron to input data. Finally, positive and irrelevant contributions in the saliency map are merged into a new feature. Numerous experimental results demonstrate that the MC-feature can greatly strengthen the slight intra-class difference in SEI and provides a possibility of interpretation of CNN.

show abstract

Robust Speaker Identification Algorithms and Results in Noisy Environments

Cited by 6 publications

References 11 publications

Acoustic Anomaly Detection of Mechanical Failures in Noisy Real-Life Factory Environments

Acoustic Anomaly Detection of Mechanical Failures in Noisy Real-Life Factory Environments

CASA-based speaker identification using cascaded GMM-CNN classifier in noisy and emotional talking conditions

A Novel Data-Driven Specific Emitter Identification Feature Based on Machine Cognition

Contact Info

Product

Resources

About