Mel-Weighted Single Frequency Filtering Spectrogram for Dialect Identification

Kethireddy, Rashmi; Kadiri, Sudarsana Reddy; Alku, Paavo; Gangashetty, Suryakanth V.

doi:10.1109/access.2020.3020506

Cited by 10 publications

(3 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The experimental results demonstrate strong performance with an accuracy of 81.26%; these results were attained by using their techniques on a common dataset of 8 English accents. Similar works include [23,24,25,26].…”

Section: Related Workmentioning

confidence: 98%

Recognition of Arabic Accents From English Spoken Speech Using Deep Learning Approach

Habbash,

Mnasri,

Alghamdi

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Accents, or changes in how different people speak the same word/sentence in the same language, pose substantial communication issues in most spoken languages. This is a well-known fact, but how does the accent of one language affect learning/speaking another? In this paper, we look at how Arab accents influence the English language. To that end, we built a deep machine-learning system for Arabic accent recognition that was learned from an in-house English speech database of four Arabic accents collected from Jordan, Iraq, Saudi Arabia, and Tunisia. The proposed system employs Mel spectrograms of an English-spoken paragraph to train an LSTM neural network to recognize the accent in each sound signal. Although the collected data was extremely difficult to learn due to the presence of both males and females and fluent speakers in each class, the proposed system could recognize speakers with various accents by up to 79%. This answers the study's main question, demonstrating that speakers with an Arabic accent have their way of speaking English, which varies by country. As a result, if trained on appropriate and adequate data, the proposed system can also be used to recognize accents in any language.

show abstract

Section: Related Workmentioning

confidence: 98%

Recognition of Arabic Accents From English Spoken Speech Using Deep Learning Approach

Habbash,

Mnasri,

Alghamdi

et al. 2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…At the same time, North Sami dialects have been identified from audio by training several models, kNNs, SVMs, RFs, CRFs, and LSTM, based on extracted features (Kakouros et al, 2020). Kethireddy et al (2020) use Mel-weighted SFF spectrogram to detect spoken Arabic dialects. Mel spectograms are also used by Draghici et al (2020).…”

Section: Related Workmentioning

confidence: 99%

Finnish Dialect Identification: The Effect of Audio and Text

Hämäläinen¹,

Alnajjar²,

Partanen³

et al. 2021

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

Finnish is a language with multiple dialects that not only differ from each other in terms of accent (pronunciation) but also in terms of morphological forms and lexical choice. We present the first approach to automatically detect the dialect of a speaker based on a dialect transcript and transcript with audio recording in a dataset consisting of 23 different dialects. Our results show that the best accuracy is received by combining both of the modalities, as text only reaches to an overall accuracy of 57%, where as text and audio reach to 85%. Our code, models and data have been released openly on Github and Zenodo.

show abstract

“…The SFF method was shown to provide better spectral features, such as harmonics, resonances (Chennupati et al, 2019;Pannala et al, 2016), and time-domain features, such as glottal closure instances and voice-onset time (VOT) (Kadiri and Yegnanarayana, 2017;Nellore et al, 2017). Inspired by the advantages of SFF, mel filter-bank energies derived from SFF (MFBE-SFF) were investigated with an SVM classifier in our previous studies (Kethireddy et al, 2020a), which showed promising results in identifying dialects compared to conventional STFT representations, such as the mel-spectrogram and MFCCs. In extension to the preliminary studies (Kethireddy et al, 2020a), this study proposes to derive four different feature representations: namely (1) SFF spectrogram (SPEC-SFF), (2) single frequency filtered cepstral coefficients (SFFCCs), (3) mel filter-bank energies derived from SFF spectrum (MFBE-SFF), and (4) mel-frequency cepstral coefficients derived from SFF spectrum (MFCC-SFF).…”

Section: Introductionmentioning

confidence: 99%

Deep neural architectures for dialect classification with single frequency filtering and zero-time windowing feature representations

Kethireddy

Kadiri

Gangashetty

2022

The Journal of the Acoustical Society of America

Self Cite

View full text Add to dashboard Cite

The goal of this study is to investigate advanced signal processing approaches [single frequency filtering (SFF) and zero-time windowing (ZTW)] with modern deep neural networks (DNNs) [convolution neural networks (CNNs), temporal convolution neural networks (TCN), time-delay neural network (TDNN), and emphasized channel attention, propagation and aggregation in TDNN (ECAPA-TDNN)] for dialect classification of major dialects of English. Previous studies indicated that SFF and ZTW methods provide higher spectro-temporal resolution. To capture the intrinsic variations in articulations among dialects, four feature representations [spectrogram (SPEC), cepstral coefficients, mel filter-bank energies, and mel-frequency cepstral coefficients (MFCCs)] are derived from SFF and ZTW methods. Experiments with and without data augmentation using CNN classifiers revealed that the proposed features performed better than baseline short-time Fourier transform (STFT)-based features on the UT-Podcast database [Hansen, J. H., and Liu, G. (2016). "Unsupervised accent classification for deep data fusion of accent and language information," Speech Commun. 78, 19-33]. Even without data augmentation, all the proposed features showed an approximate improvement of 15%-20% (relative) over best baseline (SPEC-STFT) feature. TCN, TDNN, and ECAPA-TDNN classifiers that capture wider temporal context further improved the performance for many of the proposed and baseline features. Among all the baseline and proposed features, the best performance is achieved with single frequency filtered cepstral coefficients for TCN (81.30%), TDNN (81.53%), and ECAPA-TDNN (85.48%). An investigation of data-driven filters, instead of fixed mel-scale, improved the performance by 2.8% and 1.4% (relatively) for SPEC-STFT and SPEC-SFF, and nearly equal for SPEC-ZTW. To assist related work, we have made the code available ([Kethireddy, R., and Kadiri, S. R. (2022). "Deep neural architectures for dialect classification with single frequency filtering and zero-time windowing feature representations," https://github.com/r39ashmi/e2e_dialect (Last viewed

show abstract

Mel-Weighted Single Frequency Filtering Spectrogram for Dialect Identification

Abstract: The second author would like to thank the Academy of Finland (Projects 312490 and 330139) for supporting his stay in Finland as a Postdoctoral researcher.

Cited by 10 publications

References 33 publications

Recognition of Arabic Accents From English Spoken Speech Using Deep Learning Approach

Recognition of Arabic Accents From English Spoken Speech Using Deep Learning Approach

Finnish Dialect Identification: The Effect of Audio and Text

Deep neural architectures for dialect classification with single frequency filtering and zero-time windowing feature representations

Contact Info

Product

Resources

About