GTCC-based BiLSTM deep-learning framework for respiratory sound classification using empirical mode decomposition

Jayalakshmy, S.; Sudha, Gnanou Florence

doi:10.1007/s00521-021-06295-x

Cited by 21 publications

(6 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In a different scenario where the target was to classify bird sounds, in [ 40 ], GTCC coupled with a Support Vector Machine (SVM) also outperformed MFCC and Linear Prediction Cepstrum Coefficients (LPCC). More recently, GTCC have also been used on a deep-learning framework consisting of a Bidirectional Long Short-Term Memory (BiLSTM) to classify respiratory sounds [ 41 ].…”

Section: Related Workmentioning

confidence: 99%

Analysis and Acoustic Event Classification of Environmental Data Collected in a Citizen Science Project

Bonet-Solà

Vidaña-Vila

Alsina-Pagès

2023

IJERPH

View full text Add to dashboard Cite

Citizen science can serve as a tool to obtain information about changes in the soundscape. One of the challenges of citizen science projects is the processing of data gathered by the citizens, to obtain conclusions. As part of the project Sons al Balcó, authors aim to study the soundscape in Catalonia during the lockdown due to the COVID-19 pandemic and afterwards and design a tool to automatically detect sound events as a first step to assess the quality of the soundscape. This paper details and compares the acoustic samples of the two collecting campaigns of the Sons al Balcó project. While the 2020 campaign obtained 365 videos, the 2021 campaign obtained 237. Later, a convolutional neural network is trained to automatically detect and classify acoustic events even if they occur simultaneously. Event based macro F1-score tops 50% for both campaigns for the most prevalent noise sources. However, results suggest that not all the categories are equally detected: the percentage of prevalence of an event in the dataset and its foregound-to-background ratio play a decisive role.

show abstract

Section: Related Workmentioning

confidence: 99%

Analysis and Acoustic Event Classification of Environmental Data Collected in a Citizen Science Project

Bonet-Solà

Vidaña-Vila

Alsina-Pagès

2023

IJERPH

View full text Add to dashboard Cite

show abstract

“…The feature extraction process considers Mel-Frequency Cepstral Coefficient (MFCC) [10] and Gamma Tone Frequency Cepstral Coefficient (GTCC) [3] feature extraction from the voice signals. With respect to each feature extraction techniques separately, the extracted features from the voice input are analyzed by evaluating spectral fluctuation, sharpness and roughness of the voice data.…”

Section: Feature Extractionmentioning

confidence: 99%

Ffr+dn: A Novel Deep Learning Enabled Vocal Feature Analysis for Classification of Parkinson’s Disease

Sharanyaa¹,

Sambath²

2023

INDJCSE

View full text Add to dashboard Cite

Parkinson's disease (PD) is the unique forms of disorder get impacted by non-functioning neurodegenerative parts. Vocal features plays vital role in determining the PD detection in the early stages. Analyzing the vocal patterns in-depth towards the feature extraction is important. Audio data is the complex form of data having numerous peaks of threshold points that determine the covariate information about the input signal. The proposed approach is focused on deep feature analysis of vocal patterns with spectrogram and cepstral features. The input raw vocal signals are preprocessed where signals are segmented into frames and sampled into normalized frequency, channeled into two different phases of analysis. The first phase extract the features of the vocal signals using Mel Frequency Cepstral Co-efficient (MFCC) and Gammatone Frequency Cepstral Co-efficient (GFCC). The feature extracted samples are intended to calculate Sharpness, Roughness and Fluctuation strength of the input signals. The statistical measures captured using the feature maps are formulated and tested for regression with CatBoost regression algorithm (CBR). The second phase of operation performs converting voice signal into spectrogram. Further the spectrogram converted signals are separated into training images and testing images and processed with Deep Convolution Neural Network (DCNN). The predicted labels from both the phases of operations are further cross validated with N-times of repeated testing using Dual Perceptron Neural Network (DPNN). The Novel Feature Fused Regression (FFR) mapped Deep Network (DN) (FFR+DN) architecture is formed to acquire the uniqueness of the input data. The performance measure is formulated using Accuracy, Sensitivity, Specificity and MCC. The proposed approach is achieved with accuracy of 94% and compared with various existing state-of-art approaches.

show abstract

“…Gammatone Cepstral Coefficients (GTCCs) have continued to be effective even though MFCCs have gained importance in audio data recognition in the last few years (Jayalakshmy and Sudha, 2021). GTCC feature extraction is similar to MFCC feature extraction.…”

Section: Gammatone Cepstral Coefficients (Gtcc)mentioning

confidence: 99%

“…In the next step, Gammatone filters were applied to the best combination IMF features and Gammatone cepstral coefficients (GTCC) were calculated. results show that the proposed GTCC of the third IMF component applied to the clustered BiLSTM framework outperforms the competing Convolutional Neural Network classification method in terms of accuracy, specificity, and sensitivity (Jayalakshmy and Sudha, 2021). In Kutlu and Karaca's study, feature selection was done using Mel frequency cepstral coefficients (MFCC) and gammatone cepstral coefficients (GTCC) methods.…”

Section: Introductionmentioning

confidence: 99%

Vücut Seslerinden Bölge Tanımlanması için İdeal Kayıt Süresinin Belirlenmesinde MFCC ve GTCC Özniteliklerinin Etkisinin Karşılaştırılması

Balli¹

2022

EJOSAT

View full text Add to dashboard Cite

İnsan vücudunun durumu hakkında bilgi almak için yapılabilecek en hızlı yöntemlerden birisi vücut seslerini analiz etmektir. Seslerin dijital ortama aktarılabilmesi bu analizi kolaylaştırmaktadır. Bu çalışmada kalp, akciğer ve karın bölgelerinden alınan ses verilerinden bölge tespiti yapılmıştır. Eğitimde 12 kişiden alınan 4000 örnekleme frekansına sahip 20s lik veriler kullanılmıştır. Veriler 9 farklı saniyede incelenmiştir. Her bir saniye için tüm veriler bölünmüş ve eğitim için hazırlanmıştır. MFCC ve GTCC kullanılarak öznitelikler çıkarılmış ve bu öznitelikler CNN modelinde eğitilmiştir. MFCC ve GTCC katsayılarının sonuçlar üzerindeki etkisi kıyaslanmıştır. Eğitimde en iyi sonuç %98 ile 1,5 saniyelik kayıtlardan alınan MFCC katsayısından, validationlarda ise en iyi sonuç %85 ile 1 saniyelik kayıtların MFCC katsayılarından elde edilmiştir. Genel validation sonuçlarına bakıldığında MFCC sonuçlarının daha başarılı olduğu görülmüştür.

show abstract

GTCC-based BiLSTM deep-learning framework for respiratory sound classification using empirical mode decomposition

Cited by 21 publications

References 23 publications

Analysis and Acoustic Event Classification of Environmental Data Collected in a Citizen Science Project

Analysis and Acoustic Event Classification of Environmental Data Collected in a Citizen Science Project

Ffr+dn: A Novel Deep Learning Enabled Vocal Feature Analysis for Classification of Parkinson’s Disease

Vücut Seslerinden Bölge Tanımlanması için İdeal Kayıt Süresinin Belirlenmesinde MFCC ve GTCC Özniteliklerinin Etkisinin Karşılaştırılması

Contact Info

Product

Resources

About