GFCC based discriminatively trained noise robust continuous ASR system for Hindi language

Dua, Mohit; Aggarwal, Rajesh; Biswas, Mantosh

doi:10.1007/s12652-018-0828-x

Cited by 37 publications

(11 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The performance is reported for various experimental approaches such as acoustic modelling context, modifying number of mixtures in the GMM, various optimized features with HMM, GMM-HMM, variations in the language models, using different discriminative learning techniques i.e., GFCC optimized by DE with MMI, MPE, with clean and noise corrupted speech for various types of noises of changing SNR levels. The DE optimized GFCC coupled with MPE discriminative learning techniques with triphone language models outperformed all other model combinations in clean and noise affected conditions [16].…”

Section: Literature Surveymentioning

confidence: 94%

Performance of Isolated and Continuous Digit Recognition System using Kaldi Toolkit

2019

ijrte

View full text Add to dashboard Cite

A digit recognition system is built for recognizing the sequence of digits through 0-9. The system is experimented with speech corpus created in the room environment. The acoustic information to feature representation is achieved using PLP and MFCC features. The system initially utilized the conventional GMM-HMM framework, state of the art hybrid classifier with varied number of states to complete the speech recognition task, i.e., the system is first trained and tested using Monophone models, and system’s recognition accuracy is then evaluated using Triphone Models: Triphone1 models, which was later followed by Triphones2 models and Triphones3 Models. The Ngram Language model is used for both Monophone and Triphone training. The system performance is evaluated with the use of MFCC and PLP parameterisation techniques on Kaldi toolkit. The system performance is evaluated using metrics word error rate (WER) and Word Recognition Accuracy (WRA). The proposed system can be utilized for building speech applications

show abstract

Section: Literature Surveymentioning

confidence: 94%

Performance of Isolated and Continuous Digit Recognition System using Kaldi Toolkit

2019

ijrte

View full text Add to dashboard Cite

show abstract

“…Table 6 lists all the extracted features along with their statistical measures of mean and standard deviation (STD) for method I (DWT) and method II (EMD). We extracted time domain [ 38 , 39 , 40 , 41 , 42 , 43 , 44 , 45 ], spectral [ 46 , 47 ], fractal and chaos [ 48 , 49 ], chroma [ 50 , 51 ], cepstral [ 52 ], and texture features [ 53 ] and analyzed them statistically.…”

Section: Methodsmentioning

confidence: 99%

Expert Hypertension Detection System Featuring Pulse Plethysmograph Signals and Hybrid Feature Selection and Reduction Scheme

Khan

Aziz

Akram

et al. 2021

Sensors

View full text Add to dashboard Cite

Hypertension is an antecedent to cardiac disorders. According to the World Health Organization (WHO), the number of people affected with hypertension will reach around 1.56 billion by 2025. Early detection of hypertension is imperative to prevent the complications caused by cardiac abnormalities. Hypertension usually possesses no apparent detectable symptoms; hence, the control rate is significantly low. Computer-aided diagnosis based on machine learning and signal analysis has recently been applied to identify biomarkers for the accurate prediction of hypertension. This research proposes a new expert hypertension detection system (EHDS) from pulse plethysmograph (PuPG) signals for the categorization of normal and hypertension. The PuPG signal data set, including rich information of cardiac activity, was acquired from healthy and hypertensive subjects. The raw PuPG signals were preprocessed through empirical mode decomposition (EMD) by decomposing a signal into its constituent components. A combination of multi-domain features was extracted from the preprocessed PuPG signal. The features exhibiting high discriminative characteristics were selected and reduced through a proposed hybrid feature selection and reduction (HFSR) scheme. Selected features were subjected to various classification methods in a comparative fashion in which the best performance of 99.4% accuracy, 99.6% sensitivity, and 99.2% specificity was achieved through weighted k-nearest neighbor (KNN-W). The performance of the proposed EHDS was thoroughly assessed by tenfold cross-validation. The proposed EHDS achieved better detection performance in comparison to other electrocardiogram (ECG) and photoplethysmograph (PPG)-based methods.

show abstract

“…In this study, we propose a method for monitoring the transformer status by classifying over-, normal-, and under-voltage levels based on the acoustic signal of the transformer operating in various noisy environments. In particular, we focus on the audible frequency band, which can be recognized by humans from the acoustic signal; that is, the acoustic signal is measured within the audible frequency, and the measured acoustic signal is converted into a Mel Spectrogram (MS) [ 21 , 22 , 23 ]. To design the classification model, we exploit the idea of the U-Net [ 24 ] encoder layers to extract and express the important features from the acoustic signal.…”

Section: Introductionmentioning

confidence: 99%

“…To analyze the acoustic signal, the Linear Prediction Cestrum Coefficient (LPCC) [ 21 ], Gammatone Filter Cepstral Coefficient (GFCC) [ 22 ], and Mel Frequency Cestrum Coefficient (MFCC) [ 23 ] are widely used in the time–frequency domain. LPCC can represent the acoustic signal as the frequency for a certain time without signal distortion.…”

Section: Introductionmentioning

confidence: 99%

Power Transformer Voltages Classification with Acoustic Signal in Various Noisy Environments

Kim

Lee

2022

Sensors

View full text Add to dashboard Cite

Checking the stable supply voltage of a power distribution transformer in operation is an important issue to prevent mechanical failure. The acoustic signal of the transformer contains sufficient information to analyze the transformer conditions. However, since transformers are often exposed to a variety of noise environments, acoustic signal-based methods should be designed to be robust against these various noises to provide high accuracy. In this study, we propose a method to classify the over-, normal-, and under-voltage levels supplied to the transformer using the acoustic signal of the transformer operating in various noise environments. The acoustic signal of the transformer was converted into a Mel Spectrogram (MS), and used to classify the voltage levels. The classification model was designed based on the U-Net encoder layers to extract and express the important features from the acoustic signal. The proposed approach was used for its robustness against both the known and unknown noise by using the noise rejection method with U-Net and the ensemble model with three datasets. In the experimental environments, the testbeds were constructed using an oil-immersed power distribution transformer with a capacity of 150 kVA. Based on the experimental results, we confirm that the proposed method can improve the classification accuracy of the voltage levels from 72 to 88 and to 94% (baseline to noise rejection and to noise rejection + ensemble), respectively, in various noisy environments.

show abstract

GFCC based discriminatively trained noise robust continuous ASR system for Hindi language

Cited by 37 publications

References 36 publications

Performance of Isolated and Continuous Digit Recognition System using Kaldi Toolkit

Performance of Isolated and Continuous Digit Recognition System using Kaldi Toolkit

Expert Hypertension Detection System Featuring Pulse Plethysmograph Signals and Hybrid Feature Selection and Reduction Scheme

Power Transformer Voltages Classification with Acoustic Signal in Various Noisy Environments

Contact Info

Product

Resources

About