Transformers have become the architecture of choice for learning long-range dependencies, yet their adoption in hyperspectral imaging (HSI) is still emerging. We reviewed more than 300 papers published up to 2025 and present the first end-to-end survey dedicated to Transformer-based HSI classification. The study categorizes every stage of a typical pipeline—pre-processing, patch or pixel embedding, positional encoding, spatial–spectral feature extraction, multi-head self-attention variants, skip connections, and loss design—and contrasts alternative design choices with the unique spatial–spectral properties of HSI. We map the field’s progress against persistent obstacles: scarce labeled data, extreme spectral dimensionality, computational overhead, and limited model explainability. Finally, we outline a research agenda prioritizing valuable public data sets, lightweight on-edge models, illumination and sensor shifts robustness, and intrinsically interpretable attention mechanisms. Our goal is to guide researchers in selecting, combining, or extending Transformer components that are truly fit for purpose for next-generation HSI applications.

In this paper, we present a speaker-independent dysarthric speech recognition system, with a focus on evaluating the recently released Speech Accessibility Project (SAP-1005) dataset, which includes speech data from individuals with Parkinson’s disease (PD). Despite the growing body of research in dysarthric speech recognition, many existing systems are speaker-dependent and adaptive, limiting their generalizability across different speakers and etiologies. Our primary objective is to develop a robust speaker-independent model capable of accurately recognizing dysarthric speech, irrespective of the speaker. Additionally, as a secondary objective, we aim to test the cross-etiology performance of our model by evaluating it on the TORGO dataset, which contains speech samples from individuals with cerebral palsy (CP) and amyotrophic lateral sclerosis (ALS). By leveraging the Whisper model, our speaker-independent system achieved a CER of 6.99% and a WER of 10.71% on the SAP-1005 dataset. Further, in cross-etiology settings, we achieved a CER of 25.08% and a WER of 39.56% on the TORGO dataset. These results highlight the potential of our approach to generalize across unseen speakers and different etiologies of dysarthria.

Automatic Speech Recognition (ASR) holds immense potential to provide an effective interface for assistive technologies, but its performance remains unsatisfactory for people with speech impairments such as dysarthria. Existing ASR systems struggle to accurately recognize dysarthric speech due to the significant speaker variability in dysarthric speech and the scarcity of dysarthric datasets. In this study, we propose a two-phase adaptation pipeline based on the Conformer architecture that leverages typical speech to transfer to individualized ASR models for dysarthric speakers. ASR performance is evaluated for isolated words and continuous sentences, yielding an average Word Error Rate of 21.5% on the UASpeech dataset and 12.7% on the TORGO dataset. Selectively freezing decoder layers was more often successful than selectively freezing encoder layers, suggesting that optimal performance is achieved by focusing the adaptation on the acoustic information contained in the encoder.

The design and optimization of piezoelectric transducer (PT) are one of the most challenging and critical stages in acoustic power transfer (APT). In this paper, the equivalent circuit of the PT is obtained by electromechanical analogy, and it is found that the system efficiency has nothing to do with the shape of the PT. However, the volume parameters (sectional area and thickness) of the PT greatly influence the efficiency of the system. We found that the efficiency of the system increases as the cross-sectional area of the PT increases. The thickness is just the opposite, and the system efficiency decreases as the thickness increases. In order to verify this law, the paper also established a system model in COMSOL and found that the results are highly consistent with the theoretical model. Optimizing the parameters of the PT can reduce the weight of the PT applied in life and broaden the application range for the use of the PT array.

In the computational auditory scene analysis method, the ideal ratio mask or alternatively the ideal binary mask is the key point to reconstruct the enhanced signal. The ratio mask in its Wiener filtering or its square root form is currently considered. However, this kind of ratio mask overlooked one important issue. It does not exploit the inter-channel correlation (ICC) in the noisy speech, noise, and clean speech spectra. Thus, in this paper, we first propose a novel ratio mask representation by utilizing the ICC. In this way, we adaptively reallocate the power ratio of the speech and noise during the construction of ratio mask, thus, more speech and noise components are retained and masked at the same time, respectively. Second, the channel-weight contour based on the equal loudness hearing attribute is adopted to revise this new ratio mask in each Gammatone filterbank channel. Finally, the revised ratio mask is effectively used to train a five-layer structured deep neural network. Experiments show that the proposed ratio mask performs better than the conventional ratio mask representation and other series of enhancement algorithms in terms of speech quality, intelligibility, and spectral distortion under different signal to noise ratio conditions using six types of noises.

Feature extraction from retinal images is gaining popularity worldwide as many pathologies are proved having connections with these features. Automatic detection of these features makes it easier for the specialist ophthalmologists to analyse them without spending exhaustive time to segment them manually. The proposed method automatically detects the optic disc (OD) using histogram‐based template matching combined with the maximum sum of vessel information in the retinal image. The OD region is segmented by using the circular Hough transform. For detecting fovea, the retinal image is uniformly divided into three horizontal strips and the strip including the detected OD is selected. Contrast of the horizontal strip containing the OD region is then enhanced using a series of image processing steps. The macula region is first detected in the OD strip using various morphological operations and connected component analysis. The fovea is located inside this detected macular region. The proposed method achieves an OD detection accuracy over 95% upon testing on seven public databases and on our locally developed database, University of Auckland Diabetic Retinopathy database (UoA‐DR). The average OD boundary segmentation overlap score, sensitivity and fovea detection accuracy achieved are 0.86, 0.968 and 97.26% respectively.

Segmentation of retinal vessels from retinal fundus images is the key step in the automatic retinal image analysis. In this paper we propose a new unsupervised automatic method to segment the retinal vessels from retinal fundus images. Contrast enhancement and illumination correction are carried out through a series of image processing steps followed by adaptive histogram equalization and anisotropic diffusion filtering. This image is then converted to a gray scale using weighted scaling. The vessel edges are enhanced by boosting the detail curvelet coefficients. Optic disk pixels are removed before applying fuzzy C-mean classification to avoid the misclassification. Morphological operations and connected component analysis are applied to obtain the segmented retinal vessels. The performance of the proposed method is evaluated using DRIVE database to be able to compare with other state-of-art supervised and unsupervised methods. The overall segmentation accuracy of the proposed method is 95.18% which outperforms the other algorithms.

This paper recasts the problem of online secondary path modeling in the form of a statistical inverse problem. A statistical and, in particular, a Bayesian approach towards secondary path modeling is developed and the computational issues that emerge from this approach are discussed. All signals and parameters are modeled as random variables and the degree of information concerning them is coded in their probability density functions. An abstract solution is formulated in the form of a probability density function for the secondary path model. For extracting point estimates, common statistical estimation methods are investigated. It is shown that maximum likelihood estimation is not stable; however, Bayesian method of maximum a posteriori gives a reliable solution. An adaptive algorithm is then developed to compute this solution in a computationally efficient manner. This algorithm has three advantages, compared to the traditional secondary path modeling algorithms. First, it does not cause any interference with the main active noise control algorithm. Second, it does not require any additive-noise to be injected into the secondary path. Third, it does not require any off-line initiation. The convergence of the proposed algorithm is analyzed theoretically. The validity of the theoretical results is investigated by using computer simulation. Finally successful integration of the proposed algorithm into a real-time ANC system is reported.

This paper extends the existing work on the root locus analysis of FxLMS algorithm by considering secondary path modeling errors. Rules for sketching FxLMS root locus are set out. An analytic convergence condition is then derived from the root locus plot. A deliberately-misaligned secondary path model is proposed to be used as the data preparation filter in the FxLMS algorithm. The proposed filter increases the computational efficiency of the algorithm, without changing its convergence behavior. The theoretical results are verified in practice by busing an experimental system.

This paper investigates the probability of a reference minutia repeating in another sample of the same person’s fingerprint, when that probability depends only on the consistency with which a user places their finger onto a fingerprint scanner. The investigation targets cooperative users in a civilian fingerprint recognition application. A database of 800 fingerprint samples from 100 participants was collected for the purpose of simulating such a scenario. Analysis of this database showed that such users are typically sufficiently consistent in the placement of their fingers onto the scanner to ensure that there is a 0.95 probability of a reference minutia repeating in another sample of the same fingerprint. Combining multiple samples of the same fingerprint during enrolment to filter out only the most reliable reference minutiae was shown to improve this probability even further. Additional analysis showed that, as the number of reference fingerprints used increases, the number of reference minutiae remaining for recognition purposes decreases. While this trend is expected, our results indicate that this loss in the number of reference minutiae is not very significant among cooperative users.

This paper concerns about the theoretical stability of the adaptation process performed by the Filtered-x Least Mean Square (FxLMS) algorithm in active control of acoustic noise. A dynamic model for the Variance of Residual Acoustic Noise (VRAN) is developed and it is shown that the stability of this model is a sufficient condition for the stability of the adaptation process. The basic rules governing the VRAN root locus are developed, based on which an upper-bound for the adaptation step-size is derived. This upper-bound can apply to a general case with an arbitrary secondary path, unlike the traditional upper-bound used in adaptive filter theory, which was derived only for pure delay secondary paths.

This study represents a stochastic model for the adaptation process performed on adaptive control systems by the filtered‐x least‐mean‐square (FxLMS) algorithm. The main distinction of this model is that it is derived without using conventional simplifying assumptions regarding the physical plant to be controlled. This model is then used to derive a set of closed‐form mathematical expressions for formulating steady‐state performance, stability condition and learning rate of the FxLMS adaptation process. These expressions are the most general expressions, which have been proposed so far. It is shown that some previously derived expressions can be obtained from the proposed expressions as special and simplified cases. In addition to computer simulations, different experiments with a real‐time control setup confirm the validity of the theoretical findings.

This paper proposes objective quality measures adopted in speech processing for perceptual quality evaluation of audio watermarking. Different from using an auditory perception model that mimics human auditory system, objective quality measures are introduced as alternative approach to perceive the dissimilarities caused by audio watermarking. After embedding the watermark into a variety of audio signals from EBU database, we calculate the distance between the watermarked and host signals in terms of several well- developed quality measures. For correlation analysis, subjective listening tests and a commercial evaluation tool PEMO-Q are also used to grade the differences. Pearson correlation coefficients reveal that the investigated quality measures, especially weighted spectral slope (WSS) measure, correspond well with reference ratings. Moreover, quality measures run much faster than PEMO-Q. The results indicate that objective quality measures can be the perceptual quality predictors for audio watermarking.

Background states of an EEG signal describe the distinctive variations in the amplitude of the signal with respect to time. Background state detection in EEG is used to help estimate the brain growth progress in infants. Currently, background detection is mostly done manually, which is highly subjective. This paper proposes a way to automatically detect background states for preterm infants. The distribution of the amplitude vector in a 10-minute window of 2-channel preterm neonatal EEG signal is analysed, and the mean and standard deviations of the amplitudes in log-space are used as features in a linear discriminant analysis based classifier. The results are compared with existing methods of background detection. The algorithm performs well compared with the visual classification. It also shows less sensitivity to local variations the existing algorithm are suffering from.

A novel multi-channel crosstalk resistant adaptive noise cancellation (MCRANC) algorithm is presented in this paper to enhance noise carrying speech signals. The algorithm would permit locating the microphones in close proximity as it cancels out the crosstalk effect. Results have indicated that this method outperforms the commonly used techniques in the sense of SNR improvement and speech intelligibility. A SNR improvement of 17.8 dB using MCRANC keeping highly intelligible speech was achieved in our experiments versus 9.1 dB using multi-channel ANC (MANC) with far less speech quality

This paper presents a new feature for speaker identification called perceptual log area ratio (PLAR). PLAR is closely related to the log area ratio (LAR) feature. PLAR is derived from the perceptual linear prediction (PLP) rather than the linear predictive coding (LPC). The PLAR feature derived from PLP is more robust to noise than the LAR feature. In this paper, PLAR, LAR and MFCC features were tested in a Gaussian mixture model (GMM) based speaker identification system. The F-ratio feature analysis showed that the lower order PLAR and LAR coefficients are superior in classification performance to their MFCC counterparts. The text-independent, closed-set speaker identification accuracies, as tested on KING, YOHO and the down-sampled version of TIMIT databases were 98.81%, 85.29%, 97.045 % using PLAR, 97.92%, 61.76%, 94.76 % using LAR and 96.73%, 84.31%, 96.48 % using MFCC. Those results showed that PLAR is better than LAR and MFCC in both clean and noisy environments. 1.