Deep Learning Bidirectional LSTM based Detection of Prolongation and Repetition in Stuttered Speech using Weighted MFCC
Open accessThe DOI matches its Crossref record; 46 of the 46 references it cites were found in a registry or an index.
Abstract
Stuttering is a neuro-development disorder during which normal speech flow is not fluent. Traditionally Speech-Language Pathologists used to assess the extent of stuttering by counting the speech disfluencies manually. Such sorts of stuttering assessments are arbitrary, incoherent, lengthy, and error-prone. The present study focused on objective assessment to speech disfluencies such as prolongation and syllable, word, and phrase repetition. The proposed method is based on the Weighted Mel Frequency Cepstral Coefficient feature extraction algorithm and deep-learning Bidirectional Long-Short term Memory neural network for classification of stuttered events. The work has utilized the UCLASS stuttering dataset for analysis. The speech samples of the database are initially pre-processed, manually segmented, and labeled as a type of disfluency. The labeled speech samples are parameterized to Weighted MFCC feature vectors. Then extracted features are inputted to the Bidirectional-LSTM network for training and testing of the model. The effect of different hyper-parameters on classification results is examined. The test results show that the proposed method reaches the best accuracy of 96.67%, as compared to the LSTM model. The promising recognition accuracy of 97.33%, 98.67%, 97.5%, 97.19%, and 97.67% was achieved for the detection of fluent, prolongation, syllable, word, and phrase repetition, respectively.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Figures




![Fig. 6 LSTM Cell [49] .](https://figures.cochat.ai/paper-figures/7228299858ad42fc8b230a647fcbe2d7/deep-learning-bidirectional-lstm-based-detection-of-prolongation-and-repetition-figure-6.png)
![Fig. 7 Structure of an unfolded Bi-LSTM Layer [52].](https://figures.cochat.ai/paper-figures/a0f4dedc88704e0eb42ffb0a36b2df64/deep-learning-bidirectional-lstm-based-detection-of-prolongation-and-repetition-figure-7.png)








Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“Traditional assessment of stuttering depends on counting the speech disfluencies manually, such method of assessment is arbitrary, incoherent, lengthy, and error-prone. 22”
From The role of AI in the diagnosis of speech and language disorders: A systematic mapping study (2025) · cites this paper as [22]
The cited source says what the claim attributes to it.
In the citing paper
“Furthermore, an accuracy percentage of 97.19% was achieved to detect disfluent speech (word repetition), with a sensitivity of 87.5%, and a specificity of 99.37%.”
From The role of AI in the diagnosis of speech and language disorders: A systematic mapping study (2025) · cites this paper as [22]
The cited source says what the claim attributes to it.
In the citing paper
“Deep learning Bi-LSTM: Achieved in 22 study an accuracy percentage of 98.67% to detect disfluent speech (prolongation), with a sensitivity of 97.5%, and a specificity of 98.12%.”
From The role of AI in the diagnosis of speech and language disorders: A systematic mapping study (2025) · cites this paper as [22]
The cited source says what the claim attributes to it.
In the citing paper
“While detecting the disfluent speech (syllable repetition) has achieved an accuracy percentage of 97.5%, and a sensitivity of 92.5%, and a specificity of 98.12%.”
From The role of AI in the diagnosis of speech and language disorders: A systematic mapping study (2025) · cites this paper as [22]
The cited source says what the claim attributes to it.
In the citing paper
“Lastly, achieved an accuracy percentage of 97.33% to detect fluent speech, with a sensitivity of 90%, and a specificity 98.75%.”
From The role of AI in the diagnosis of speech and language disorders: A systematic mapping study (2025) · cites this paper as [22]
1 more claim about this paper
Create a free account to see every claim other papers make about this one.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Hochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318 · cited 1×
“1) Long-Short Term Memory: LSTM is a specialized Recurrent Neural Network (RNN) architecture, competent in learning long term dependencies [48].”
BackgroundBidirectional recurrent neural networks
Schuster, M., & Paliwal, K. K. (1997) · IEEE Transactions on Signal Processing · doi:10.1109/78.650093 · Cited by 10,172 · cited 8×
“RNN suffers from short-term memory, caused by vanishing gradient problem.”
BackgroundInfluentialLearning long-term dependencies with gradient descent is difficult
Bengio, Y., Simard, P., & Frasconi, P. (1994) · IEEE Transactions on Neural Networks · doi:10.1109/72.279181 · Cited by 8,542 · cited 1×
“However, the RNN is good at modeling the temporal data but suffers from the problem of short-term memory caused by vanishing gradient [33][34][35].”
BackgroundFundamentals of speech recognition
Rabiner, L. R., & Juang, B.-H. F. (1993) · Cited by 7,680
On the difficulty of training recurrent neural networks
Pascanu, R., Mikolov, T., & Bengio, Y. (2012) · arXiv · doi:10.48550/arxiv.1211.5063 · Cited by 4,310 · cited 1×
“However, the RNN is good at modeling the temporal data but suffers from the problem of short-term memory caused by vanishing gradient [33][34][35].”
BackgroundThe Vanishing Gradient Problem During Learning Recurrent Neural Nets and Problem Solutions
Hochreiter, S. (1998) · International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems · doi:10.1142/s0218488598000094 · Cited by 2,939 · cited 1×
“They are capable of learning long term dependencies [37].”
BackgroundDigital Processing of Speech Signals
Ackroyd, M. H. (1979) · Electronics and Power · doi:10.1049/ep.1979.0173 · Cited by 2,391
Spoken Language Processing: A Guide to Theory, Algorithm, and System Development
Huang, X., Acero, A., Hon, H.-W., & Reddy, R. (2001) · Cited by 1,593
The Performance of LSTM and BiLSTM in Forecasting Time Series
Siami-Namini, S., Tavakoli, N., & Namin, A. S. (2019) · 2019 IEEE International Conference on Big Data (Big Data) · doi:10.1109/bigdata47090.2019.9005997 · Cited by 1,574 · cited 1×
“Hence gives better classification accuracy as compared to one-way LSTM [53].”
ResultGradient Flow in Recurrent Nets: The Difficulty of Learning LongTerm Dependencies
(2009) · A Field Guide to Dynamical Recurrent Networks · doi:10.1109/9780470544037.ch14 · Cited by 1,246
Research works printed but not linked: 5
- [5][5] S. Gupta, R. S. Shukla, and R. K. Shukla, 'Literature survey and review of techniques used for automatic assessment of Stuttered Speech,' Int. J. Manag. Technol. Eng., vol. IX, no. X, pp. 229-240, 2019. -
- [8][8] K. M. Ravikumar, R. Rajagopal, and H. C. Nagaraj, 'An Approach for Objective Assessment of Stuttered Speech Using MFCC Features,' vol. 9, no. 1, pp. 19-24, 2009. -
- [17][17] J. Pálfy, 'Analysis of Dysfluencies by Computational Intelligence,' Information Sci. Technol. Bull. ACM Slovakia, vol. 6, no. 2, pp. 45-58, 2014. -
- [39][39] P. Howell, S. Davis, and J. Bartrip, 'Europe PMC Funders Group The UCLASS archive of stuttered speech,' vol. 52, no. 2, pp. 556-569, 2010. -
- [51][51] A. Graves, N. Jaitly, and A. Mohamed, 'Hybrid speech recognition with deep bidirectional LSTM Alex Graves , Navdeep Jaitly and Abdelrahman Mohamed University of Toronto Department of Computer Science 6 King ' s College Rd . Toronto , M5S 3G4 , Canada,' pp. 273278, 2013. -
Cited by
Mel Frequency Cepstral Coefficient and its Applications: A Review
Abdul, Z. K., & Al-Talabani, A. K. (2022) · IEEE Access · doi:10.1109/access.2022.3223444 · Cited by 527
Tang, Z. L., Bo, L., Liu, X., & Wei, D. (2021) · Applied Intelligence · doi:10.1007/s10489-021-02504-1 · Cited by 40
“Engineers extract the acoustic features (mel-frequency cepstral coefficients, log-filterbank energy) of speech and feed them into the LSTM network for natural language modeling [36].”
BackgroundSystematic Review of Machine Learning Approaches for Detecting Developmental Stuttering
Barrett, L., Hu, J., & Howell, P. (2022) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2022.3155295 · Cited by 49
Świetlicka, I., Kuniszyk–Jóźkowiak, W., & Świetlicki, M. (2022) · Sensors · doi:10.3390/s22010321 · Cited by 29
“Wavelet Transform (DWT) or Perceptual Linear Prediction (PLP), are sensitive to noise and achieve excellent results for clear speech [35,47,48].”
BackgroundComputational Intelligence-Based Stuttering Detection: A Systematic Review
Alnashwan, R., Alhakbani, N., Al-Nafjan, A., Almudhi, A., & Al-Nuwaiser, W. M. (2023) · Diagnostics · doi:10.3390/diagnostics13233537 · Cited by 29
“Moreover, Gupta et al. [21] employed the BLSTM model, achieving an impressive overall classification accuracy of 96.67% in detecting various types of stuttered events.”
BackgroundMethodInfluentialAbubakar, M., Mujahid, M., Kanwal, K., Iqbal, S., Asghar, M. N., & Alaulamie, A. A. (2024) · IEEE Access · doi:10.1109/access.2024.3429343 · Cited by 14
“Sequential input processing and capturing extended dependencies are strengths of LSTM [21].”
BackgroundDeep analysis of MFCC and MEL spectrogram features to recognize and classify stuttered speech
Banerjee, N., Sethi, N., & Borah, S. (2025) · Multimedia Tools and Applications · doi:10.1007/s11042-025-20881-4 · Cited by 8
“14-dimensional 14-dimensional weighted MFCC features from the UCLASS database were taken as input in Bi-LSTM network to detect Prolongation and repetition (syllable, word, and phrase repetition) types of stuttering [ 27 ].”
Binh, N. H., Hai, D. V., Dat, B. T., Chau, H. N., & Cuong, N. Q. (2022) · International Journal of Advanced Computer Science and Applications · doi:10.14569/ijacsa.2022.0131088 · Cited by 6
“Moreover, Recurrent Neural Networks (RNNs), especially the Long Short-Term Memories (LSTMs), have been experimented with to be able to capture the information of the long sequence [53-56].”
BackgroundThe role of AI in the diagnosis of speech and language disorders: A systematic mapping study
Tbaishat, D., Al-Shafei, R., & Odeh, M. (2025) · DIGITAL HEALTH · doi:10.1177/20552076251379769 · Cited by 4
Nonlinear Dynamic System Identification of ARX Model for Speech Signal Identification
Pattanaik, R. K., Mohanty, M. N., Mohapatra, S. K., & Pattanayak, B. K. (2023) · Computer Systems Science and Engineering · doi:10.32604/csse.2023.029591 · Cited by 4
Multi-modal approach for COVID-19 detection using coughs and self-reported symptoms
Nguyen-Trong, K., & Nguyen-Hoang, K. (2022) · Journal of Intelligent & Fuzzy Systems · doi:10.3233/jifs-222863 · Cited by 8
A mixed generative model of auditory word repetitionPreprint
Sajid, N., Holmes, E., Costa, L. D., Price, C. J., & Friston, K. J. (2022) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2022.01.20.477138 · Cited by 6
“However, in (Gupta et al., 2020) it learns contextualised speech using a bidirectional LSTM that introduces temporal scheduling of exchanges.”
BackgroundMethodGupta, S. (2021) · Indian Journal of Science and Technology · doi:10.17485/ijst/v14i5.2276 · Cited by 3
“67% and performs better than other models (26).”
BackgroundSelf-citationDetection and Classification of Stuttering from Text
Moghimi, F., & Yazdi, M. (2024) · International Symposium on Telecommunications · doi:10.1109/ist64061.2024.10843520 · Cited by 3
“With the growing interest in artificial neural networks and advancements in deep learning, newer techniques have been developed, leading to higher accuracy as more data became available[1, 20].”
InfluentialAnonymization of Stuttered Speech -- Removing Speaker Information while Preserving the Utterance
Hintz, J., Bayerl, S. P., Sinha, Y., Ghosh, S., Schubert, M., Stober, S., Riedhammer, K., & Siegert, I. (2023) · Symposium on Security and Privacy in Speech Communication (SPSC) · doi:10.21437/spsc.2023-7 · Cited by 6
“A prerequisite for these applications are techniques to automatically process and detect disordered speech, which is a sub-field of pathological speech research with a focus on developing systems for the automatic processing, screening, and monitoring of non-normal speech recordings [3].”
de Almeida, R. J. S., Souza, D. Y., Oliveira, L. P., Correia, D. V., Pinheiro, S. R. N. B., & Sousa, E. S. D. S. (2024) · doi:10.5753/sbcas.2024.2319 · Cited by 1
Comparison of Machine Learning Algorithms for Detection of Stuttering in Speech
Gudlavalleti, S., Devi, P. S., Lakka, R., Kuchanpally, R., & Dudekula, S. S. (2025) · Springer proceedings in mathematics & statistics · doi:10.1007/978-3-031-51338-1_30 · Cited by 1
Hybrid Approach to Detect Prolonged Speech Segments
Drakshayini, K. B., & Anusuya, M. A. (2023) · International Journal of Engineering and Advanced Technology · doi:10.35940/ijeat.d4106.0412423 · Cited by 1
Jokar, A. H. R., Khoshnevis, A., & Saber-Moghadam, R. (2026) · Journal of Fluency Disorders · doi:10.1016/j.jfludis.2026.106222
Stuttered Speech Recognition And Classification Using Enhanced Kamnan Filter And Neural Network
Vaidianathan, B., Arulselvi, S., & Karthik, B. (2021) · International Conference on Computing, Communication and Control System (I3CAC) · doi:10.4108/eai.7-6-2021.2308659
“[1 ] The technology that is used to convert an audio speech into text information is called as Automatic Speech Recognition (ASR) system.”
MethodComparison of Machine Learning Algorithms for Detection of Stuttering in Speech
Devi, P. S. (2023) · Advances in Computational Sciences and Technology · doi:10.37622/acst/16.1.2023.45-53
Drakshayini, K. B., Anusuya, M. A., Vani, H. Y., & Chayadevi, M. L. (2025) · SN Computer Science · doi:10.1007/s42979-025-03958-7
Energy-efficient scalable memory LSTM cell with approximate feature processing units
Sindhu, R., & Arunachalam, V. (2025) · Results in Engineering · doi:10.1016/j.rineng.2025.108517 · Cited by 1
Banerjee, N., Sethi, N., & Borah, S. (2025) · International Conference on Emerging Research in Electronics, Computer Science and Technology (ICERECT) · doi:10.1109/icerect65215.2025.11376013
“Sakshi Gupta et al. developed a Bidirectional LSTM model aimed at identifying fluent, prolongation, syllable, word, and phrase repetition in Stuttered Speech by employing Weighted MFCC [12].”
Stutter classification on the SEP-28k dataset: Insights from ResNet-18 and BiLSTM architechures
Patil, V. R., Myakalmardi, T. S., Netravathi, K., Chikkamath, S., Nirmala, S. R., & Budihal, S. V. (2026) · AIP conference proceedings · doi:10.1063/5.0340315
Stammering Detection with Streaming TDNN and Log Mel Acoustic Features
Kaviya, L., Ravichandran, M., Sivasankaran, S., Jayasree, P., Sakthiyavathi, M., & Keerthana, R. (2026) · International Conference on Data Science for Cyber-Physical Systems Resilience using Advanced Applications (ICDCA) · doi:10.1109/icdca69396.2026.11620420
“The bidirectional design broadens the contextual comprehension by using speech chunks not only from the past but also from the future, thus helping to get the higher precision of classification [20].”
Metadata
- Authors
- Gupta, S., S., R., K., R., & Verma, R.
- Affiliations
- Invertis University (IN)
- King Khalid University (SA)
- Venue
- International Journal of Advanced Computer Science and Applications
- Published
- 2020
- Volume
- 11
- Issue
- 9
- ISSN
- 2156-5570 (ISSN-L)
- 2158-107X
- OpenAlex
- W3090973479
- Cited by
- 26
- References
- 45
- Citations by year
- 2026: 3 · 2025: 6 · 2024: 3 · 2023: 5 · 2022: 6 · 2021: 3
- FWCI
- 1.39
- Type
- journal article
- Language
- en
- Open access
- diamond
- doi.org
- Licence
- cc-by
- Locations
- International Journal of Advanced Computer Science and Applications (publishedVersion, cc-by)
- International Journal of Advanced Computer Science and Applications (submittedVersion, cc-by)
- Topics
- Social Sciences › Psychology › Clinical Psychology › Stuttering Research and Treatment
- Physical Sciences › Computer Science › Artificial Intelligence › Speech Recognition and Synthesis
- Health Sciences › Health Professions › Speech and Hearing › Voice and Speech Disorders
- Keywords
- stuttering detection, disfluency, prolongation, syllable repetition, word repetition, phrase repetition, bidirectional LSTM
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.