Najim Dehak

Dr. Najim Dehak


Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
14 December 2020 in IEEE Signal Processing Letters
Verified
Journal Article
Closed access
NCNanxin Chen
Nanxin Chen
SWShinji Watanabe
Shinji Watanabe
JVJesus Antonio Villalba
Jesus Antonio Villalba
PZPiotr Zelasko
Piotr Zelasko
NDNajim Dehak
Najim Dehak
Very deep transformers outperform conventional bi-directional long short-term memory networks for automatic speech recognition (ASR) by a significant margin. However, being autoregressive models, their computational complexity is still a prohibitive factor in their deployment into production systems. To amend this problem, we study two different non-autoregressive transformer structures for ASR: Audio-Conditional Masked Language Model (A-CMLM) and Audio-Factorized Masked Language Model (A-FMLM). When training these frameworks, the decoder input tokens are randomly replaced by special mask tokens. Then, the network is optimized to predict the masked tokens by taking both the unmasked context tokens and the input speech into consideration. During inference, we start from all masked tokens and the network iteratively predicts missing tokens based on partial results. A new decoding strategy is proposed as an example, which starts from the most confident predictions to the rest. Results on Mandarin (AISHELL), Japanese (CSJ), English (LibriSpeech) benchmarks show promising results to train such a non-autoregressive network for ASR. Especially in AISHELL, the proposed method outperformed the Kaldi ASR system and matched the performance of the state-of-the-art autoregressive transformer with 7 × speedup.

29 September 2021 in IEEE Transactions on Information Forensics and Security
Verified
Journal Article
Closed access
SJSonal Joshi
Sonal Joshi
JVJesus Villalba
Jesus Villalba
PZPiotr Zelasko
Piotr Zelasko
LMLaureano Moro-Velazquez
Laureano Moro-Velazquez
NDNajim Dehak
Najim Dehak
Adversarial examples are designed to fool the speaker recognition (SR) system by adding a carefully crafted human-imperceptible noise to the speech signals. Posing a severe security threat to state-of-the-art SR systems, it becomes vital to deep-dive and study their vulnerabilities. Moreover, it is of greater importance to propose countermeasures that can protect the systems against these attacks. Addressing these concerns, we first investigated how state-of-the-art x-vector based SR systems are affected by white-box adversarial attacks, i.e., when the adversary has full knowledge of the system. x-Vector based SR systems are evaluated against white-box adversarial attacks common in the literature like fast gradient sign method (FGSM), basic iterative method (BIM)–a.k.a. iterative-FGSM–, projected gradient descent (PGD), and Carlini-Wagner (CW) attack. To mitigate against these attacks, we investigated four pre-processing defenses which do not need adversarial examples during training. The four pre-processing defenses–viz. randomized smoothing, DefenseGAN, variational autoencoder (VAE), and Parallel WaveGAN vocoder (PWG) are compared against the baseline defense of adversarial training. Performing powerful adaptive white-box adversarial attack (i.e., when the adversary has full knowledge of the system, including the defense), our conclusions indicate that SR systems were extremely vulnerable under BIM, PGD, and CW attacks. Among the proposed pre-processing defenses, PWG combined with randomized smoothing offers the most protection against the attacks, with accuracy averaging 93% compared to 52% in the undefended system and an absolute improvement >90% for BIM attacks with $L_\infty >0.001$ and CW attack.

7 June 2022 in IEEE/ACM Transactions on Audio, Speech, and Language Processing
Verified
Journal Article
Closed access
SBSaurabhchand Bhati
Saurabhchand Bhati
JVJesus Villalba
Jesus Villalba
PZPiotr Zelasko
Piotr Zelasko
LMLaureano Moro-Velazquez
Laureano Moro-Velazquez
NDNajim Dehak
Najim Dehak
Typically, unsupervised segmentation of speech into the phone- and wordlike units are treated as separate tasks and are often done via different methods which do not fully leverage the inter-dependence of the two tasks. Here, we unify them and propose a technique that can jointly perform both, showing that these two tasks indeed benefit from each other. Recent attempts employ self-supervised learning, such as contrastive predictive coding (CPC), where the next frame is predicted given past context. However, CPC only looks at the audio signal's frame-level structure. We overcome this limitation with a segmental contrastive predictive coding (SCPC) framework to model the signal structure at a higher level, e.g., phone level. A convolutional neural network learns frame-level representation from the raw waveform via noise-contrastive estimation (NCE). A differentiable boundary detector finds variable-length segments, which are then used to optimize a segment encoder via NCE to learn segment representations. The differentiable boundary detector allows us to train frame-level and segment-level encoders jointly. Experiments show that our single model outperforms existing phone and word segmentation methods on TIMIT and Buckeye datasets. Finally, we use SCPC to extract speech features at the segment level rather than at the uniformly spaced frame level (e.g., 10 ms) and produce variable rate representations that change according to the contents of the utterance. We lower the feature extraction rate from the typical 100 Hz to 14.5 Hz on average while still outperforming the hand-crafted features such as MFCC on the linear phone classification task.

19 January 2022 in Applied Sciences
Verified
Journal Article
Open access
PDF file
AOAlex S. Ozbolt
Alex S. Ozbolt
LMLaureano Moro-Velazquez
Laureano Moro-Velazquez
ILIoan A Lina
Ioan A Lina
ABAnkur A. Butala
Ankur A. Butala
NDNajim Dehak
Najim Dehak
Diagnosing Parkinson’s Disease (PD) necessitates monitoring symptom progression. Unfortunately, diagnostic confirmation often occurs years after disease onset. A more sensitive and objective approach is paramount to the expedient diagnosis and treatment of persons with PD (PwPDs). Recent studies have shown that we can train accurate models to detect signs of PD from audio recordings of confirmed PwPDs. However, disparities exist between studies and may be caused, in part, by differences in employed corpora or methodologies. Our hypothesis is that unaccounted covariates in methodology, experimental design, and data preparation resulted in overly optimistic results in studies of PD automatic detection employing sustained vowels. These issues include record-wise fold creation rather than subject-wise; an imbalance of age between the PwPD and control classes; using too small of a corpus compared to the sizes of feature vectors; performing cross-validation without including development data; and the absence of cross-corpora testing to confirm results. In this paper, we evaluate the influence of these methodological issues in the automatic detection of PD employing sustained vowels. We perform several experiments isolating each issue to measure its influence employing three different corpora. Moreover, we analyze if the perceived dysphonia of the speakers could be causing differences in results between the corpora. Results suggest that each independent methodological issue analyzed has an effect on classification accuracy. Consequently, we recommend a list of methodological steps to be considered in future experiments to avoid overoptimistic or misleading results.

8 August 2022 in IEEE Journal of Selected Topics in Signal Processing
Verified
Journal Article
Closed access
JCJaejin Cho
Jaejin Cho
JVJesus Villalba
Jesus Villalba
LMLaureano Moro-Velazquez
Laureano Moro-Velazquez
NDNajim Dehak
Najim Dehak
In recent studies, self-supervised pre-trained models tend to outperform supervised pre-trained models in transfer learning. In particular, self-supervised learning of utterance-level speech representation can be used in speech applications that require discriminative representation of consistent attributes within an utterance: speaker, language, emotion, and age. Existing frame-level self-supervised speech representation, e.g., wav2vec, can be used as utterance-level representation with pooling, but the models are usually large. There are also self-supervised learning techniques to learn utterance-level representation. One of the most successful is a contrastive method, which requires negative sampling: selecting alternative samples to contrast with the current sample (anchor). However, this does not ensure that all the negative samples belong to classes different from the anchor class without labels. This paper applies a non-contrastive self-supervised method to learn utterance-level embeddings. We adapted DIstillation with NO labels (DINO) from computer vision to speech. Unlike contrastive methods, DINO does not require negative sampling. We compared DINO to x-vector trained in a supervised manner. When transferred to speaker verification, speech emotion recognition (SER), and Alzheimer's disease detection, DINO outperformed x-vector. We studied the influence of several aspects during transfer learning such as dividing the fine-tuning process into steps, chunk lengths, or augmentation. Fine-tuning the last affine layers first and then the whole network surpassed fine-tuning all at once. Using shorter chunk lengths, although they generate more diverse inputs, did not necessarily improve performance, implying speech segments at least with a specific length are required for better performance per application. Augmentation was helpful in SER.