Speech enhancement is crucial both for human and machine listening applications. Over the last decade, the use of deep learning for speech enhancement has resulted in tremendous improvement over the ...classical signal processing and machine learning methods. However, training a deep neural network is not only time-consuming; it also requires extensive computational resources and a large training dataset. Transfer learning, i.e. using a pretrained network for a new task, comes to the rescue by reducing the amount of training time, computational resources, and the required dataset, but the network still needs to be fine-tuned for the new task. This paper presents a novel method of speech denoising and dereverberation (SD&D) on an end-to-end frozen binaural anechoic speech separation network. The frozen network requires neither any architectural change nor any fine-tuning for the new task, as is usually required for transfer learning. The interaural cues of a source placed inside noisy and echoic surroundings are given as input to this pretrained network to extract the target speech from noise and reverberation. Although the pretrained model used in this paper has never seen noisy reverberant conditions during its training, it performs satisfactorily for zero-shot testing (ZST) under these conditions. It is because the pretrained model used here has been trained on the direct-path interaural cues of an active source and so it can recognize them even in the presence of echoes and noise. ZST on the same dataset on which the pretrained network was trained (homo-corpus) for the unseen class of interference, has shown considerable improvement over the weighted prediction error (WPE) algorithm in terms of four objective speech quality and intelligibility metrics. Also, the proposed model offers similar performance provided by a deep learning SD&D algorithm for this dataset under varying conditions of noise and reverberations. Similarly, ZST on a different dataset has provided an improvement in intelligibility and almost equivalent quality as provided by the WPE algorithm.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
This paper presents a novel sound event detection (SED) system for rare events occurring in an open environment. Wavelet multiresolution analysis (MRA) is used to decompose the input audio clip of 30 ...seconds into five levels. Wavelet denoising is then applied on the third and fifth levels of MRA to filter out the background. Significant transitions, which may represent the onset of a rare event, are then estimated in these two levels by combining the peak-finding algorithm with the K-medoids clustering algorithm. The small portions of one-second duration, called 'chunks' are cropped from the input audio signal corresponding to the estimated locations of the significant transitions. Features from these chunks are extracted by the wavelet scattering network (WSN) and are given as input to a support vector machine (SVM) classifier, which classifies them. The proposed SED framework produces an error rate comparable to the SED systems based on convolutional neural network (CNN) architecture. Also, the proposed algorithm is computationally efficient and lightweight as compared to deep learning models, as it has no learnable parameter. It requires only a single epoch of training, which is 5, 10, 200, and 600 times lesser than the models based on CNNs and deep neural networks (DNNs), CNN with long short-term memory (LSTM) network, convolutional recurrent neural network (CRNN), and CNN respectively. The proposed model neither requires concatenation with previous frames for anomaly detection nor any additional training data creation needed for other comparative deep learning models. It needs to check almost 360 times fewer chunks for the presence of rare events than the other baseline systems used for comparison in this paper. All these characteristics make the proposed system suitable for real-time applications on resource-limited devices.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
The ongoing COVID-19 pandemic is caused by SARs-CoV-2. The virus is transmitted from person to person through droplet infections i.e. when infected person is in close contact with another person. In ...January 2020, first report of detection of SARS-CoV-2 in faeces, has made it clear that human wastewater might contain this virus. This may illustrate the probability of environmentally facilitated transmission, mainly the sewage, however, environmental conditions that could facilitate faecal oral transmission is not yet clear. We used existing Pakistan polio environment surveillance network to investigate presence of SARs-CoV-2 using three commercially available kits and E-Gene detection published assay for surety and confirmatory of positivity. A Two-phase separation method is used for sample clarification and concentration. An additional high-speed centrifugation (14000Xg for 30 min) step was introduced, prior RNA extraction, to increase viral RNA yield resulting a decrease in Cq value. A total of 78 wastewater samples collected from 38 districts across Pakistan, 74 wastewater samples from existing polio environment surveillance sites, 3 from drains of COVID-19 infected areas and 1 from COVID 19 quarantine center drainage, were tested for presence of SARs-CoV-2. 21 wastewater samples (27%) from 13 districts turned to be positive on RT-qPCR. SARs-COV-2 RNA positive samples from areas with COVID 19 patients and quarantine center strengthen the findings and use of wastewater surveillance in future. Furthermore, sequence data of partial ORF 1a generated from COVID 19 patient quarantine center drainage sample also reinforce our findings that SARs-CoV-2 can be detected in wastewater. This study finding indicates that SARs-CoV-2 detection through wastewater surveillance has an epidemiologic potential that can be used as supplementary system to monitor viral tracking and circulation in cities with lower COVID-19 testing capacity or heavily populated areas where door-to-door tracing may not be possible. However, attention is needed on virus concentration and detection assay to increase the sensitivity. Development of highly sensitive assay will be an indicator for virus monitoring and to provide early warning signs.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
We present a novel unsupervised fall detection system that employs the collected acoustic signals (footstep sound signals) from an elderly person׳s normal activities to construct a data description ...model to distinguish falls from non-falls. The measured acoustic signals are initially processed with a source separation (SS) technique to remove the possible interferences from other background sound sources. Mel-frequency cepstral coefficient (MFCC) features are next extracted from the processed signals and used to construct a data description model based on a one class support vector machine (OCSVM) method, which is finally applied to distinguish fall from non-fall sounds. Experiments on a recorded dataset confirm that our proposed fall detection system can achieve better performance, especially with high level of interference from other sound sources, as compared with existing single microphone based methods.
•A data description method is applied to detect falls by collected acoustic signals (footstep sound signals) from normal activities.•A spectral subtraction based binaural dereverberation method is applied to reduce the sound reverberation for pre-processing.•A time-frequency (TF) mask based blind source separation technique is applied to separate the sound signal with the interferences of noises.•One class support vector machine method is employed for the data description of the Mel-frequency cepstral coefficient features to distinguish fall and non-fall sounds.
A series of carbamothioyl-furan-2-carboxamide derivatives were synthesized using a one-pot strategy. Compounds were obtained in moderate to excellent yields (56-85%). Synthesized derivatives were ...evaluated for their anti-cancer (HepG2, Huh-7, and MCF-7 human cancer cell lines) and anti-microbial potential. Compound p-tolylcarbamothioyl)furan-2-carboxamide showed the highest anti-cancer activity at a concentration of 20 μg/mL against hepatocellular carcinoma, with a cell viability of 33.29%. All compounds showed significant anti-cancer activity against HepG2, Huh-7, and MCF-7, while indazole and 2,4-dinitrophenyl containing carboxamide derivatives were found to be less potent against all tested cell lines. Results were compared with the standard drug doxorubicin. Carboxamide derivatives possessing 2,4-dinitrophenyl showed significant inhibition against all bacterial and fungal strains with inhibition zones (I.Z) in the range of 9-17 and MICs were found to be 150.7-295 μg/mL. All carboxamide derivatives showed significant anti-fungal activity against all tested fungal strains. Gentamicin was used as the standard drug. The results showed that carbamothioyl-furan-2-carboxamide derivatives could be a potential source of anti-cancer and anti-microbial agents.
A major challenge in educational technology integration is to engage students with different affective characteristics. Also, how technology shapes attitude and learning behavior is still lacking. ...Findings from educational psychology and learning sciences have gained less traction in research. The present study was conducted to examine the efficacy of a group format of an Artificial Intelligence (AI) powered writing tool for English second postgraduate students in the English academic writing context. In the present study, (N = 120) students were randomly allocated to either the equipped AI (n = 60) or non-equipped AI (NEAI). The results of the parametric test of analyzing of covariance revealed that at post-intervention, students who participated in the AI intervention group demonstrated statistically significant improvement in the scores, of the behavioral engagement (Cohen's d = .75, 95% CI 0.38, 1.12), of the emotional engagement Cohen's d = .82, 95% CI 0.45, 1.25, of the cognitive engagement, Cohen's d = .39,95% CI 0.04, .76, of the self-efficacy for writing, Cohen's d = .54, 95% CI 0.18, 0.91, of the positive emotions Cohen's d = . 44, 95% CI 0.08, 0.80, and of the negative emotions, Cohen's d = −.98, 95% CI −1.36, −0.60, compared with NEAI. The results suggest that AI-powered writing tools could be an efficient tool to promote learning behavior and attitudinal technology acceptance through formative feedback and assessment for non-native postgraduate students in English academic writing.
Artificial Intelligence, automated writing evaluation, L2 Writing, feedback, pedagogy, ESL, formative assessment.
Fall injuries and trauma-related hospitalizations are the most common causes of injury and in-hospital stay amongst the elderly population. After the age of 65, the severity and frequency of ...fall-related problems increases; the repercussions are challenging for senior citizens, caregivers, and healthcare professionals. This study aims to determine the injuries and outcomes resulting from falls in elderly patients presenting to Emergency Department of a tertiary care hospital.
A cohort study design was used. All elderly patients aged ≥ 60 years who visit the Emergency Department with a history of a fall as a primary complaint presenting to the ED of a tertiary care hospital in Karachi, Pakistan were included. A purposive sampling strategy was used to enroll 318 patients from August 2021 to February 2022. The outcome was risk of mortality. Each individual was followed for 90 days to study the outcome. A multivariable logistic regression was applied to check the association between the outcome and covariates. Crude and adjusted risk ratios were reported. A p-value ≤ 0.05 was considered significant.
Of the 318 participants, 265 (83.3%) were fall injury patients with comorbidities. More than half of the patients in both groups were female 32 (60.4%) & 146 (55.1%). Eyeglasses were used by most of the fall patients both without and with comorbidities 21 (39.6%) & 152 (57.4%) p 0.018. There were multiple reasons for a fall including imbalance/dizziness, which was reported by one third of participants in both groups 15 (28.3%) & 77 (29.1%) followed by a fall from stairs/steps/escalator 15 (28.3%) & 44 (16.6%) p 0.005. At the end of one month, of those who had a comorbidity 20 (7.5%) expired. The risk of mortality among fall related injuries in elderly patients who were more than 80 years was 1.48 times (95% CI: 1.20-2.10) more likely when compared to those patients who were younger than 80 years.
Efforts should be made to improve management of the underlying etiology of falls to prevent them in future. The factors that contribute to falls should be identified. Strategies and interventions should be planned to mitigate this risk of fall in elderly to improve their quality of life.
Electricity theft presents a significant financial burden to utility companies globally, amounting to trillions of dollars annually. This pressing issue underscores the need for transformative ...measures within the electrical grid. Accordingly, our study explores the integration of block chain technology into smart grids to combat electricity theft, improve grid efficiency, and facilitate renewable energy integration. Block chain's core principles of decentralization, transparency, and immutability align seamlessly with the objectives of modernizing power systems and securing transactions within the electricity grid. However, as smart grids advance, they also become more vulnerable to attacks, particularly from smart meters, compared to traditional mechanical meters. Our research aims to introduce an advanced approach to identifying energy theft while prioritizing user privacy, a critical aspect often neglected in existing methodologies that mandate the disclosure of sensitive user data. To achieve this goal, we introduce three distributed algorithms: lower-upper decomposition (LUD), lower-upper decomposition with partial pivoting (LUDP), and optimized LUD composition (OLUD), tailored specifically for peer-to-peer (P2P) computing in smart grids. These algorithms are meticulously crafted to solve linear systems of equations and calculate users' "honesty coefficients," providing a robust mechanism for detecting fraudulent activities. Through extensive simulations, we showcase the efficiency and accuracy of our algorithms in identifying deceitful users while safeguarding data confidentiality. This innovative approach not only bolsters the security of smart grids against energy theft, but also addresses privacy and security concerns inherent in conventional energy-theft detection methods.