Search for: All records

Speech enhancement through denoising is essential for maintaining signal intelligibility and quality in biometric speaker verification pipelines that operate in acoustically adverse conditions. Despite the proliferation of deep learning (DL) architectures for speech denoising, simultaneously optimizing noise attenuation, perceptual fidelity, and speaker-identity preservation remains an open problem. We address this gap by benchmarking three architecturally distinct DL-based enhancement models—Wave-U-Net, CMGAN, and U-Net—on three independent, domain-diverse corpora (SpEAR, VPQAD, and Clarkson) that the models never encountered during training and by introducing commercial-grade VeriSpeak speaker-verification scores as a biometric evaluation dimension absent from prior comparative studies. Our experiments reveal a clear three-way trade-off: U-Net achieves the highest signal-to-noise ratio (SNR) gains (+61.44% on SpEAR, +67.05% on VPQAD, +235.3% on Clarkson) but sacrifices naturalness; CMGAN yields the best perceptual evaluation of speech quality (PESQ) values (3.33, 1.35, and 2.50, respectively), favoring listening-comfort applications; and Wave-U-Net delivers the strongest biometric fidelity (VeriSpeak improvements of +11.63%, +30.22%, and +29.24%) while offering competitive perceptual quality. These results highlight that model selection must be driven by the target deployment scenario and provide actionable guidance for improving biometric verification robustness under real-world noise. 

The accurate and timely diagnosis of neurodegenerative disorders such as Parkinson’s disease, Alzheimer’s disease, and major depressive disorder critically depends on real-time monitoring and precise interpretation of authentic neurotransmitter (NT) signal dynamics in complex biological fluids (CBFs), including cerebrospinal fluid. These NT dynamics are governed by both the type and concentration of NTs present in the CBFs. However, current biosensors face significant limitations in sensitivity and selectivity, thereby hindering reliable estimation (detection and quantification) of NTs. Though nanomaterials and bioenzymes have been utilized to modify sensor interfaces for enhanced performance, issues like signal convolution, electrode fouling, and inter-NT crosstalk persist. Objectives: This review aims to evaluate and synthesize current research on the use of artificial intelligence (AI), particularly machine learning (ML), pattern recognition (PR), and deep learning (DL), to improve the automated detection and quantification of neurotransmitters from complex biological fluids. Design: A systematic review of 33 peer-reviewed studies was conducted, focusing on the integration of AI methods in neurotransmitter estimation. The review includes an analysis of commonly studied NTs, the methodologies for their detection, data acquisition techniques, and the AI algorithms applied for signal processing and interpretation. Results: The studies reviewed demonstrate that AI-based approaches have shown considerable potential in overcoming traditional biosensor limitations by effectively deconvoluting complex, multiplexed NT signals. These techniques allow for more accurate NT estimation in real-time monitoring scenarios. The review categorizes AI methodologies by their application and performance in NT signal analysis. Conclusions: AI-enhanced NT monitoring represents a promising direction for advancing diagnostic and therapeutic capabilities in neurodegenerative diseases. Despite current challenges, such as sensor stability and NT interaction complexity, AI integration, particularly in applications like closed-loop deep brain stimulation (CLDBS), offers significant potential for more effective and personalized treatments. 

Large fingerprint datasets, while important for training and evaluation, are time‐consuming and expensive to collect and require strict privacy measures. Researchers are exploring the use of synthetic fingerprint data to address these issues. This article presents a novel approach for generating synthetic fingerprint images (both spoof and live), addressing concerns related to privacy, cost, and accessibility in biometric data collection. Our approach utilizes conditional StyleGAN2‐ADA and StyleGAN3 architectures to produce high‐resolution synthetic live fingerprints, conditioned on specific finger identities (thumb through little finger). Additionally, we employ CycleGANs to translate these into realistic spoof fingerprints, simulating a variety of presentation attack materials (e.g., EcoFlex, Play‐Doh). These synthetic spoof fingerprints are crucial for developing robust spoof detection systems. Through these generative models, we created two synthetic datasets (DB2 and DB3), each containing 1500 fingerprint images of all 10 fingers with multiple impressions per finger, and including corresponding spoofs in eight material types. The results indicate robust performance: our StyleGAN3 model achieves a Fréchet inception distance (FID) as low as 5, and the generated fingerprints achieve a true acceptance rate (TAR) of 99.47% at a 0.01% false acceptance rate (FAR). The StyleGAN2‐ADA model achieved a TAR of 98.67% at the same 0.01% FAR. We assess fingerprint quality using standard metrics (NFIQ2, MINDTCT), and notably, matching experiments confirm strong privacy preservation, with no significant evidence of identity leakage, confirming the strong privacy‐preserving properties of our synthetic datasets. 

In this work, we simultaneously detected and predicted the concentration levels of serotonin (SE) and dopamine (DA) neurotransmitters (NTs) for in vitro mixtures, with measurements obtained using conventional glassy carbon electrodes (CGCEs) and differential pulse voltammetry (DPV). The NTs were estimated by deconvolving the multiplexed signals of both NTs using Principal Component Analysis with Gaussian Process Regression (PCA-GPR) and Partial Least Squares with Gaussian Process Regression (PLS-GPR), both with exponential–isotropic kernels. The average testing accuracies of estimation using PCA-GPR for DA alone, SE alone and their mixture (DA–SE) were 87.6%, 88.1%, and 96.7%, respectively. Using PLS-GPR, the testing accuracies of estimation for DA alone, SE alone, and their mixture (DA–SE) were 87.3%, 83.8%, and 95.1%, respectively. Furthermore, we explored methods of reducing the procedural complexity in estimating the NTs by finding reduced subsets of features for accurately detecting and predicting their concentrations. The reduced subsets of features found in the oxidation potential windows of the NTs improved the testing accuracy of the estimation of DA–SE to 97.4%. We thus believe that reducing complexity has the potential to increase the detection and prediction accuracies of NT measurements for practical clinical uses such as deep brain stimulation.