Objective speech intelligibility metrics (OIMs) are widely used in various fields, including public ser- vice announcements. These metrics do not directly predict the intelligibility of a speech (defined as the ratio of understandable words in an audio sample), but produce values that tend to monotonically increase with intelligibility. Several mapping functions, typically logistic models, are applied to raw objective scores to produce accurate predictions. However, there exists no standard methodology for choosing the best mapping curve, therefore, researchers tend to reuse curves originally meant for other datasets and OIMs. This research ap- plies a method called Akaike Information Criterion (AIC), specifically developed for model selection, to existing candidate models as well as new ideas based on simple heuristics. Afterwards, the models are evaluated using AIC. The new criterion affirmed the logistic mapping functions chosen for the objective intelligibility metrics STOI and MIKNN, and highlighted alternative models for the SIIB and SIIBgauss. However, with too few listening conditions on the dataset, strong inferences could not be easily made from the data.

Respiratory Rate (RR) is a vital health indicator, especially in infant monitoring, where early detection of abnormalities or variabilities in RR is crucial. Traditionally, the respiratory rate is extracted using contact-based methods, which, although reliable, can be quite intrusive and stressful for long-term monitoring. This study explores the potential of real-time remote RR monitoring on inexpensive hardware, by comparing three motion-based methods of extracting RR from RGB-camera feed: Pixel Intensity Changes (PIC), Optical Flow (OF), and Eulerian Video Magnification (EVM). The three algorithms were benchmarked using the public AIR-125 dataset, which features videos of infants in various positions, with a focus on their accuracy and computational intensity. The results show that the PIC algorithm slightly outperformed the other two algorithms in both accuracy and computational complexity. However, none of the algorithms managed to replicate the performance of the study which initially proposed the dataset as a benchmark.

This study investigates the performance of two objective speech quality metrics, Perceptual Evaluation of Speech Quality (PESQ) and Virtual Speech Quality Objective Listener (ViSQOL), in predicting human-rated speech quality scores, which are essential for telecommunication systems' Quality of Experience (QoE). These metrics have been validated using a limited number of languages due to the insufficiency of labeled data with human-rated scores. This research focuses on the applicability of PESQ and ViSQOL in Turkish and Korean, two languages that were not part of the validation data for calibrating these metrics.

The experiment used English as the baseline language for comparison, and the results showed that Turkish samples had higher average ViSQOL scores, with the difference being statistically significant compared to the English samples. Furthermore, Turkish male speakers had the highest correlation between PESQ and ViSQOL scores, and ViSQOL rated speech higher than PESQ, especially under babble noise degradations. Future research should focus on extending this study by exploring biases across additional metrics and languages, while also constructing a dataset with labeled subjective scores for more languages to improve the calibration of these metrics.

Clear communication in public address systems is essential, especially in environments where safety or information clarity is critical. Speech intelligibility is often assessed using objective intelligibility metrics (OIMs), which predict intelligibility through mathematical models. These metrics perform well in non-highly reverberant conditions but face challenges in highly reverberant environments and with non-European languages like Mandarin. This study examines the performance of three intrusive OIMs—ESTOI, HASPI, and SIIB\textsubscript{Gauss}—in two aspects: (1) how these metrics perform under different reverberation conditions for English, using STIPA as a reference, and (2) how robust these metrics are by comparing the variances of scores between Mandarin and English. The results show that the variances of predicted scores by the test metrics are equal between Mandarin and English. HASPI, ESTOI, and SIIB\textsubscript{Gauss} demonstrate similar performance across a broader range of reverberation conditions (from a T60 of 0.05s to 7s) for English, contradicting the theory that most intrusive intelligibility metrics struggle with severe reverberation conditions \cite{galbrun_speech_2016}. The findings highlight the need for further research to evaluate potential biases in OIMs and their performance across languages. Incorporating listening tests could provide a more solid examination of these metrics under diverse conditions for different languages.

Heart rate (HR) is a critical indicator of an individual’s health, serving as a key metric for detecting potential cardiac issues. This paper explores a method for real-time heart rate measurement using RGB camera footage, aimed at general health monitoring. The proposed method utilizes a convolutional neural network (CNN) to generate a 3D mesh of the subjects’ facial features. The movement over time of the points in this mesh is used to compute a signal that captures the small pulsatile movements corresponding to the mechanical motion of blood being pumped through the veins. This signal is filtered, and motion sources are separated using principal component analysis (PCA). The most periodic component, the one with the highest frequency, is considered to correspond to the heart rate, and it’s frequency is used to estimate the heart rate. The proposed method is tested using the ECG-Fitness dataset, characterized by challenging environmental conditions such as significant subject motion and dim lighting conditions. Experimental results demonstrate the method’s capability for real-time applications, though further enhancements are needed to improve robustness under difficult environmental conditions.

Assessing speech intelligibility for individuals with dysarthria is critical for understanding the severity of motor speech disorders and evaluating speech therapy interventions. Traditional subjective as- sessments, while effective, are resource-intensive and prone to bias, which highlights the need for reliable objective measures. This study investi- gates the applicability of MIKNN (Mutual Infor- mation with K-Nearest Neighbors) as an objective speech intelligibility measure for dysarthric speech, by comparing objective intelligibility scores with subjective ratings. Unlike its proven effective- ness with neurotypical speech, the performance of objective measures on atypical speech, such as dysarthria, remains under-explored. The study compares MIKNN with state-of-the-art measures, including P-STOI and P-ESTOI, using the UA- Speech dataset. Key challenges addressed in- clude adapting MIKNN to handle the temporal and spectral variability inherent in dysarthric speech. The results demonstrate that while MIKNN offers promising correlations with subjective scores, it is outperformed by P-STOI and P-ESTOI.

The measurement of the heart rate (HR) is of vitalimportance in modern medicine. Advancements in medical technology have resulted in a myriad of techniques to measure and analyze these bio-signals, and the advent of telemedicine and the post-COVID-19 world has placed greater emphasis on contact-free measurement tools.

Previous works have explored several methods of contact-free heart rate measurement. Remote Photo-Plethysmography(rPPG) measures HR from RGB camera streams by detecting and analyzing the frequency of “micro-blushes” in the skin corresponding to pulse; this method can also be used to estimate SpO2. Eulerian video magnification instead attempts to detectsubtle micro-movements caused by the pulse to measure HR aswell as RR. Ballistocardiography tracks longitudinal movements of feature points to estimate HR. However, there is little research on applying these (typically computationally-intensive) algorithms in the context of real-time, low-performance embedded systems.

This paper evaluates the computational requirements of algorithms used in extracting bio-signals using an RGB camera. It focuses on rPPG, a camera based method for extracting heart rate from video. Several rPPG implementations are tested on standard hardware and benchmarked on real-time performance and memory. The experiments were conducted on publicly available datasets. Additionally, region-of-interest selection algorithms are also compared. These results are of much use in developing embedded devices for remote monitoring of bio-signals and provide some insight into algorithms viable for use in real-time contexts.

Reverberation is a key aspect when designing the interior of buildings, and must be carefully considered in the context of the function of the room. Defined by the reverberation time (RT), it is known to have a big influence on the intelligibility and quality of audio in closed spaces.
In this work, we investigate the relationship between the RT and explore the feasibility of using multilayer perceptron (MLP) networks to create an estimator for the RT by using the values of objective measures as input features. We investigate five measures in particular: the Perceptual Evaluation of Speech Quality (PESQ), Virtual Speech Quality Objective Listener (ViSQOL) and its extension focused on audio (ViSQOLAudio), and the Short-time Objective Intelligibility Measure (STOI) and its extension ESTOI.
We create a 3-layer MLP network that estimates the RT with a mean absolute error of 0.144 on our simulated RIR test sets and 0.196 on our real RIR test set.

In building design, it is important to consider certain materials for certain acoustical properties. Specifically, the time it takes for an audio signal to decrease in volume by 60 dB is important. This can be estimated with Sabine's and Eyring's formula's, which both make use of the average absorption coefficient of the materials in a room. This absorption coefficient indicates how much of the original audio signal is absorbed into the material. However, measuring these absorption coefficients for a material is difficult and time consuming.
In this study, a machine learning approach is used to estimate the absorption coefficients, by using the room impulse response in combination with the layout of a room. A room impulse response is the characterizing sound of a room. These two pieces of data are processed through a convolutional neural network and a multilayer perceptron, respectively, and combined to make the final prediction of absorption coefficients. A novel approach of simulating the data is used, and a real dataset is used in conjunction with the simulations to use a state-of-the-art regression loss function made for domain adaptation. The results show that the machine learning approach still has a large error compared to using Eyring's formula, and that machine learning is not yet a viable option to use instead of conventional methods.

The accurate simulation of Room Impulse Responses (RIRs) is important in a variety of applications in acoustics such as automatic speech recognition, speech enhancement, and architectural acoustic design. While objective metrics for evaluating RIRs have been researched extensively, the subjective perceptual accuracy of the simulations is largely overlooked. This paper seeks to address this gap, designing a subjective testing methodology for evaluating the perceptual accuracy of simulated RIRs. A framework is proposed that combines the ABX testing methodology with a modified Multiple Stimuli with Hidden Reference and Anchor (MUSHRA) approach, measuring attributes such as clarity, warmth, environment, and reverbera- tion. The study involved 50 participants evaluating audio samples convolved with both real and simulated RIRs. Results seem to indicate that participants could reliably distinguish between real and simulated RIRs, with perceptual differences observed in the “clarity” and “reverberation” attributes. The findings suggest that current simulation methods for RIRs do not fully capture the perceptual aspects of acoustic environments.

Estimating reverberation time (RT60) accurately is crucial for enhancing the acoustic quality of various environments as it decides how you feel the sound fades away subjectively. Traditional methods, such as Sabine's equation, require extensive prior knowledge and assume ideal conditions, limiting their practicality. To address these limitations, this paper explores the application of convolutional neural networks (CNNs) enhanced with an encoder architecture based on transformer mechanisms for blind RT60 estimation. The proposed model leverages simulated and real-world datasets, incorporating environmental noise to improve robustness. Results indicate that the CNN-Encoder model achieves superior performance, with a mean squared error (MSE) as low as 0.0006 seconds for pure room impulse responses (RIRs) and 0.0011 seconds under +30dB signal-to-noise ratio (SNR) conditions. It also demonstrates potential in practical usage achieving an MSE of 0.0282 seconds under audio recordings. This approach offers a significant reduction in estimation error compared to the CNN-only architecture, demonstrating the potential for improved acoustic parameter estimation in varied environments. Future work will focus on further optimizing the model for real-world applications and reducing computational complexity while maintaining high accuracy.

Dutch State-of-the-art Automatic Speech Recognition (ASR) systems do not perform equally well for different speaker groups. Existing metrics to quantify this bias rely on demographic metadata, which is often unavailable. Recent advances in the field use machine learning to find groups of similar speakers instead. However, its black-box nature obscures the interpretability of resulting groups. This paper proposes an interpretable approach to bias discovery by clustering speakers based on acoustic and prosodic features. Different feature subsets were compared in their ability to find performance disparities in five ASR systems for two separate speaking styles. Results show that these feature sets can uncover bias approaching known disparities between demographic groups. While the effectiveness per feature set differed between the speaking styles, the most successful ones found significant disparities between clusters with diverse demographic compositions.

Automatic Speech Recognition (ASR) systems have become an integral part of daily lives. Despite their widespread use, these systems can exhibit biases that express themselves in the differences in their accuracy and performance across different demographic groups. Methods quantifying these biases have been developed. This paper investigates the relationship between bias and the acoustic characteristics of speakers. By examining various acoustic embeddings, derived from models like wav2vec 2.0 and XLSR, we aim to identify which embeddings correlate most strongly with bias. The findings offer insights into improving the fairness of ASRs by exploring how acoustic features influence bias in ASR systems. Future research directions include exploring isolated speech properties and extending the study to diverse linguistic contexts to deepen understanding in this area.

Automatic Speech Recognition (ASR) systems are becoming increasingly popular in this day and age. Unfortunately, due to inherent biases within these systems, performance disparities exist among specific demographic groups. Bias metrics can be used to measure this bias. Within ASR they represent a niche area that has not yet been thoroughly explored. The few bias metrics that exist in literature mainly centre around the performance differences between speaker groups. This paper proposes two new bias metrics that focus not only on performance differences, but also take the base performance into account: Weighted Performance Bias (WPB) and Intergroup Weighted Performance Bias (IWPB). Although the lack of ground truth makes the results less easily interpretable, the results show similar trends within the new metrics as those defined in literature: bias is greatest among non-native Dutch speech.

This paper presents a novel approach to measuring bias in Automatic Speech Recognition (ASR) systems by proposing a metric that does not use the conventional approach of a reference group. Current methods typically measure bias through comparison with a ’norm’ or minimum error group, potentially introducing additional biases. To address this issue, this study introduces a new metric: a combination of the Groupto-Average Log Ratio and the Sum of Group Error Differences. This metric aims to provide a fair comparison by measuring performance relative to the average of the groups rather than a single reference group. Results indicate that the new metric reveals different aspects of bias not captured by traditional methods. This study contributes to the ongoing research on fairness in speech technology by challenging the existing bias metrics and proposing alternatives that might offer more equitable evaluations. Future research should explore further refinements of these metrics, and apply them across more varied datasets and environments. Ultimately, this research moves towards making ASR technologies more inclusive, ensuring that they serve all user groups equitably.

A curve is a continuously bending line with no angles that can be found anywhere in the real world, forming shapes and outlines. They are also the building blocks of historic watermarks, imprinted images on paper that may be used to identify its manufacturers. Their shapes consist of curves as bent wires are used in their production process. Often, the processing of scans of those curves may introduce gaps or a degraded quality which could be corrected by reconstructing the curves in those gaps. Curve reconstruction is a fundamental problem with many research applications, one of which is the reconstruction of curves for binarised scans of historic watermarks. In this paper, a data generation approach is proposed for the simulation of the watermark curves domain through singular automatically generated curves and human-drawn sketches which are then used along binarised watermark scans. I propose a hybrid method combining machine-learning and analytical approaches for curve reconstruction, aiming to leverage their advantages together. The method is compared to its components separately. Quantitative results against them demonstrate the superiority of the pure machine learning approach, as well as the need for more research into potentially better analytical components and a more realistic domain simulation.

Watermarks have an essential role in identifying the origins and age of specific documents. However, this is often a laborious process. One of the main issues in automatic watermark segmentation is the presence of text that obstructs it, making it difficult to properly reconstruct a watermark. Image processing and machine learning techniques face limitations, requiring time, training of data, or manual parameter selection. This research introduces a new method using wavelets transform to locate and remove text from a watermarked image, while preserving the underlying watermark. This method manages to outperform classic image processing techniques for the case when text is thicker than the watermark outline.

Feature extraction techniques for content-based image retrieval are explored, focusing on black-and-white images in the context of historical watermarks. Orthogonal moments and texture features are found to be most applicable. Seven methods are evaluated: four different orthogonal moments, Gabor features, and two novel combinations of orthogonal moments with Gabor features. Retrieval effectiveness is judged based on the precision-recall curve and mean average precision, and watermarks are considered when unchanged, rotated, sheared and both rotated and sheared. The results demonstrate that research into improving efficient grayscale image representation does not translate over to improvements with black-and-white images. As it stands, the basic Zernike moments and novel Gabor-Zernike features are most effective.

This paper tackles the problem of evaluating the task of finding similar scanned historical watermarks - small images embedded in historical paper that have been digitized to be processed on a computer - using pre-trained neural networks. This research aims to identify an efficient and accurate alternative to the traditional, time-consuming manual detection methods for finding similar watermarks. The primary issue addressed is the inefficiency of these manual methods. The evaluation focuses on finding similar watermarks for a specific query watermark, assessing the efficacy of neural networks in comparison to a prior art system that employs traditional image processing techniques. This comparison aims to determine how well these neural networks perform in the task of watermark similarity detection. The study involves a dataset of 500 labeled images tested in two distinct contexts: one using unprocessed images and another using images processed to keep only the watermark outline. The results show that pre-trained models achieve higher accuracy and time efficiency compared to the prior art system that uses image processing. These models demonstrate significant effectiveness in watermark recognition and comparison, with each network achieving over 80% accuracy for traced watermarks. EfficientNetB0 achieved 94.66%, VGG16 89.33%, ResNet50 86.67%, and InceptionV3 84%, while the prior art system gets 64,8%. These results conclude that these models are valuable tools in the field of watermark recognition and comparison.

A watermark image is a scan of a historical paper document that contains a watermark, which is a motif embedded in the paper that provides valuable information on the origins of a document. Developing tools to automatically identify watermarks can make this information more accessible to researchers. This paper focuses on one specific binarization technique, thresholding. Thresholding selects a threshold value, which is used to turn an image binary such that one color represents foreground and the other represents background. Ideally, binarization isolates the watermark’s shape by representing it as foreground, and removes unwanted information. This research compares the effectiveness of different thresholding techniques when applied to watermark images. Eight algorithms are selected from the literature, and a novel algorithm is proposed that seeks to improve on the other algorithms when applied to watermarks. The nine total algorithms are evaluated quantitatively on synthetic data, and qualitatively through a survey where participants select which algorithm appears best and rate it. The results show that there is no clear algorithm which works best for all images, however a logical adaptive approach may work marginally better than other approaches. Additionally, the presented algorithms do not adequately remove non-watermark information from the images. Further research should be conducted to analyze different binarization techniques in this context.