Speaker recognition based on MFCC and BP neural networks

Wang, Yi; Lawlor, Bob

doi:10.1109/issc.2017.7983644

Cited by 18 publications

(10 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The existing speaker modeling methods in the literature consist of spectrogram-based methods [6,7], Gaussian mixture models [8][9][10], dynamic time warping [11], vector quantization [12,13], neural networks [14,15], hidden Markov models [16,17] and so on. Mel frequency cepstral coefficients [18], linear predictive coding [19], linear predictive cepstral coefficients [20], perceptual linear predictive Fig. 3 Evolution of speaker recognition models from the traditional to the state-of-the-art methods [21], gammatone frequency cepstral coefficients [22] are some of the successful feature extraction methods for speaker recognition.…”

Section: Related Workmentioning

confidence: 99%

Speaker Identification Approach for the Post-pandemic Era of Internet of Things

Saleema

Thampi

2021

Lecture Notes in Electrical Engineering

View full text Add to dashboard Cite

With the rapid aggravation of COVID-19 pandemic, the organizations, industries, institutions, etc., are forced to rapidly adapt to social distancing by limiting physical contact and thereby limit the person to person contamination. The scenario of Internet of things in the post COVID era will be interesting indeed. Inorder to ensure public health, the social distancing and semi lockdown will continue in the foreseeable future, and therefore, the need of secure remote person authentication methods is being more and more critical especially in the Internet of things which is a multitude of networks consisting of a huge number of uniquely identifiable devices. As far as human_device authentication strategies are concerned, the one which needs less human involvement will be preferable in a post COVID-19 IoT scenario. Moreover, since the range of IoT devices may span from tiny sensors to complex machines, an authentication method which will be adaptable to each and every type of device will be more welcomed. Considering these facts, voice biometric authentication seems to be the most suitable one which can provide a balanced mix of security, adaptivity and convenience to such an advanced world of connectivity. Here, we introduce a lightweight text independent voice biometric method for IoT using extreme learning machines, and we perform a comparative analysis with a deep learning-based method of speaker identification using 3D convolutional neural networks. We have performed experimental study using different datasets and concluded that the extreme learning-based method is more suitable for IoT, considering the trade-off between the recognition accuracy and the training time requirements.

show abstract

Section: Related Workmentioning

confidence: 99%

Speaker Identification Approach for the Post-pandemic Era of Internet of Things

Saleema

Thampi

2021

Lecture Notes in Electrical Engineering

View full text Add to dashboard Cite

show abstract

“…MFCC tekniği, insan kulağının frekans bandının değişimi ile uyumlu olduğu bilinmektedir [14], [15]. Ayrıca, MFCC ses tanıma sistemlerinde diğer tekniklere göre daha başarılı olduğu kanıtlanmıştır [16], [17]. MFCCs tekniği kuru ve ıslak öksürük ayrımında faydalı bir yöntem olduğu Chatrzarrin ve ark.…”

Section: Mel-frekans Kepstral Katsayıları (Mfcc)unclassified

Solunum Hastalıkları ile İlişkili Semptom Seslerinin Sınıflandırılması

Melek¹

2021

European Journal of Science and Technology

View full text Add to dashboard Cite

ÖzCovid-19 gibi solunum yolu enfeksiyonlarının erken tespiti, hastalığın daha kolay tedavisine ve hastanın daha rahat bir süre geçirmesine yol açarak ciddi komplikasyon olasılığını azaltabilir. Öksürme ve hapşırma gibi solunum seslerinin sıklığı, şiddeti ve türü (kuru veya balgamlı), hastalığın teşhisi, tedavisi ve davranışlarının tespitinde tıp uzmanları için çıkarılabilen zengin bilgiler taşımaktadır. Bunun için, makine veya derin öğrenimine dayalı otomatik yaklaşımların geliştirilmesi oldukça önemlidir. Center for Open Science (OSFHOME), 2020 yılında güncellediği veri küme üzerine, bu alanda çalışan araştırmacıları, ses kayıtlarını kullanarak hastalık seslerinin otomatik algılanması için makine öğrenimi modelleri oluşturmaya davet etti. Veri seti, "Pfizer Digital Medicine Challenge" için oluşturulmuştur ve amacı öksürme ve hapşırma gibi seslerinin tespiti için makine öğrenimi modellerinin geliştirilmesidir. Veri seti üç parçaya ayrılmıştır; eğitim, doğrulama ve test kümeleri. Sunulan çalışmada, bu veri seti üzerine yeni bir makine öğrenimi sistemi önerildi. Eğitim, doğrulama ve test örneklerinden öznitelikler elde edildikten sonra, dört farklı sınıflandırıcının parametrelerini hesaplamak için doğrulama veri kümesi kullanıldı ve son aşamada test veri kümesi üzerine sınıflandırma gerçekleştirildi. Elde edilen sonuçlara göre, radyal tabanlı çekirdek fonksiyonlu destek vektör makine (DVM) sınıflandırıcısı solunum seslerini diğer seslere karşı, %76 civarında bir doğruluk oranıyla diğer sınıflandırıcılara göre daha başarılı sınıflandırdı.

show abstract

“…Speaker recognition based on MFCC and Back Propagation Neural Network (BPNN) is discussed in [10]. The input speech signal features are extracted by MFCC and the classification is made by BPNN.…”

Section: Introductionmentioning

confidence: 99%

Speaker Recognition System Based on Wavelet Features and Gaussian Mixture Models

Sajeer*,

Rodrigues

2019

IJEAT

View full text Add to dashboard Cite

Identification of a person’s voice from the different voices is known as speaker recognition. The speech signals of individuals are selected by means of speaker recognition or identification. In this work, an efficient method for speaker recognition is made by using Discrete Wavelet Transform (DWT) features and Gaussian Mixture Models (GMM) for classification is presented. The input speech signal features are decomposed by DWT into subband coefficients. The DWT subband coefficient features are the input for the classification. Classification is made by GMM classifier at 4, 8, 16 and 32 Gaussian component levels. Results show a better accuracy of 96.18% speaker signals using DWT features and GMM classifier

show abstract

Speaker recognition based on MFCC and BP neural networks

Cited by 18 publications

References 6 publications

Speaker Identification Approach for the Post-pandemic Era of Internet of Things

Speaker Identification Approach for the Post-pandemic Era of Internet of Things

Solunum Hastalıkları ile İlişkili Semptom Seslerinin Sınıflandırılması

Speaker Recognition System Based on Wavelet Features and Gaussian Mixture Models

Contact Info

Product

Resources

About