Article
Psychology, Multidisciplinary
Lin Lai
Summary: With the development of modern information technology, the flipped classroom teaching mode has become a hot topic in contemporary education and is being applied in various disciplines. However, this teaching mode still faces challenges such as low efficiency and lack of teacher-student interaction, leading to low student enthusiasm for learning. Thus, further testing and revision of the flipped classroom teaching mode is needed.
FRONTIERS IN PSYCHOLOGY
(2022)
Article
Chemistry, Multidisciplinary
Badriyya B. Al-onazi, Muhammad Asif Nauman, Rashid Jahangir, Muhmmad Mohsin Malik, Eman H. Alkhammash, Ahmed M. Elshewey
Summary: In recent years, data science has been widely applied in various fields such as human-computer interaction, computer gaming, mobile services, and emotion evaluation. One emerging and challenging research topic is speech emotion recognition (SER), where deep learning techniques have shown promising results.
APPLIED SCIENCES-BASEL
(2022)
Article
Chemistry, Analytical
Shouyan Chen, Mingyan Zhang, Xiaofen Yang, Zhijia Zhao, Tao Zou, Xinqi Sun
Summary: This paper discusses the applicable rules of Global-Attention and Self-Attention in SER classification construction, and proposes a new classifier model with an accuracy of 85.427% on the EMO-DB dataset.
Article
Chemistry, Multidisciplinary
Mikel de Velasco, Raquel Justo, Asier Lopez Zorrilla, Maria Ines Torres
Summary: In this study, the authors present an approach to understand how emotions are identified in spontaneous speech using computational methods and decision-making. The researchers focused on Spanish TV debates, which pose a high level of complexity and subjectivity in their annotation procedure based on human perception. They proposed a simple convolutional neural model and analyzed its decision-making process. The model showed slightly better performance than commonly used CNN architectures like VGG16, while being less resource-intensive. The researchers also visualized and analyzed the internal layer-by-layer transformations of the input spectrogram. Additionally, they proposed a class model visualization as a simple interpretation approach and evaluated its usefulness in the work.
APPLIED SCIENCES-BASEL
(2023)
Article
Psychology, Multidisciplinary
Elena Nicoladis, Chris Westbury, Cassandra Foursha-Stevenson
Summary: This study examined the impact of implicit English gender connotations on L1 English speakers' judgments of L2 French object gender, finding that English gender connotations can influence both accuracy and response time in grammatical judgments in French. This supports the idea of semantics-mediated cross-linguistic influence.
FRONTIERS IN PSYCHOLOGY
(2021)
Article
Acoustics
Juli Cebrian
Summary: This study examines the perceived similarity between English and Catalan vowels and diphthongs, finding that non-native vowels are typically perceived as instances of native categories and emphasizing the importance of contrasting native and non-native perception. The results suggest the potential of reciprocal approaches for making predictions about non-native perception and second language development.
JOURNAL OF THE ACOUSTICAL SOCIETY OF AMERICA
(2021)
Article
Health Care Sciences & Services
Ray F. Lin, Shu-Hsing Cheng, Yung-Ping Liu, Cheng-Pin Chen, Yi-Jyun Wang, Shu-Ying Chang
Summary: A preliminary study developed an artificial intelligence model aimed at discriminating the emotional valence of people living with HIV. By collecting voice clips and emotional valence values for data preprocessing and modeling, the study found that the model performed well for individuals with sufficient data.
Article
Computer Science, Artificial Intelligence
Fengkai Li, Xuan Zhang
Summary: Facial recognition, as a direct and effective biometric technology, has become a mainstream, stable, and reliable method of identification. It involves collecting biometric data and using computer processing to match templates for facial recognition.
Article
Computer Science, Information Systems
Dahiru Tanko, Fahrettin Burak Demir, Sengul Dogan, Sakir Engin Sahin, Turker Tuncer
Summary: This research aims to develop an accurate speech emotion recognition model to evaluate the effectiveness of the distance education system. The proposed model achieved a 93.40% classification accuracy using multi-level discrete wavelet transform and one-dimensional orbital local binary pattern for feature extraction, followed by feature selection using neighborhood component analysis and classification using support vector machine.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Chemistry, Physical
Beyene Hagos Aregawi, Tazeddinova Diana, Chia-Hung Su, A. S. El-Shafay, May Alashwal, Bassem F. Felemban, Mohammed Zwawi, Mohammed Algarni, Fu-Ming Wang
Summary: A model based on machine learning technique was developed to predict the molecular diffusivity of organic compounds in water at infinite dilution. The model considered diverse nonelectrolyte organic compounds and provided a comprehensive method for prediction of diffusivity at infinite dilution and temperature of 25 degrees C.
JOURNAL OF MOLECULAR LIQUIDS
(2022)
Article
Engineering, Multidisciplinary
Weixuan Hu, Shukuan Zhao
Summary: The research identifies four difficulties in human resource management: concept complexity, limited data collection, ethical issues, and evidence-based management feedback. To address these issues, AI-assisted concepts for data analysis could be considered.
INTERNATIONAL JOURNAL OF TECHNOLOGY MANAGEMENT
(2021)
Article
Psychology, Multidisciplinary
Xu Wu, Qian Zhang
Summary: This article proposes a design based on voice emotion recognition for aging intelligent home products with RBF. By combining the Hidden Markov model and the Radial Basis Function Neural Network, the recognition rate is greatly improved. Furthermore, the introduction of the dynamic optimal learning rate concept optimizes the efficiency of the network.
FRONTIERS IN PSYCHOLOGY
(2022)
Article
Computer Science, Artificial Intelligence
Turker Tuncer, Sengul Dogan, U. Rajendra Acharya
Summary: An innovative method for speech emotion recognition is proposed in this study, utilizing cryptographic structure and iterative neighborhood component analysis to achieve high classification performance. Experimental results demonstrate that the method attained high classification accuracies on multiple public databases, with potential for application in large-scale databases and healthcare settings.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Chemistry, Analytical
Giovanni Costantini, Valerio Cesarini, Pietro Di Leo, Federica Amato, Antonio Suppa, Francesco Asci, Antonio Pisani, Alessandra Calculli, Giovanni Saggio
Summary: This study analyzed the voice characteristics of Parkinson's disease patients using machine learning techniques, and compared different feature selection and classification algorithms. The results showed that both feature-based machine learning and deep learning achieved comparable results in terms of classification, with KNN, SVM, and naive Bayes classifiers performing similarly. The superiority of CFS as the best feature selector was more evident, and the selected features acted as relevant vocal biomarkers capable of differentiating healthy subjects, early untreated PD patients, and mid-advanced L-Dopa treated patients.
Article
Chemistry, Analytical
Gulmira Bekmanova, Banu Yergesh, Altynbek Sharipbay, Assel Mukanova
Summary: This article presents an emotional speech recognition method for recognizing student emotions during online exams in distance learning due to COVID-19. The method achieves an accuracy of 79.7% for the Kazakh language and can be widely applied to recognize emotions in different languages. It analyzes human speech using emotionally charged words stored in a code book to determine the presence of emotions.
Article
Environmental Sciences
Emilia Parada-Cabaleiro, Anton Batliner, Markus Schedl
Summary: Musical listening is widely used to reduce anxiety, but the acoustic properties of anxiety-reducing music have not been thoroughly studied. This study explores whether the acoustic parameters used in music emotion recognition are also suitable for identifying music with relaxing properties. The results show that when using classical Western music to reduce anxiety, tonal music should be considered and harmonicity is an appropriate indicator of relaxing music. Further research is needed to understand the role of scoring and dynamics in reducing listener distress.
INTERNATIONAL JOURNAL OF ENVIRONMENTAL RESEARCH AND PUBLIC HEALTH
(2022)
Correction
Engineering, Electrical & Electronic
Emilia Parada-Cabaleiro, Anton Batliner, Alice Baird, Bjorn Schuller
INTERNATIONAL JOURNAL OF SPEECH TECHNOLOGY
(2022)
Article
Computer Science, Information Systems
Shuo Liu, Adria Mallol-Ragolta, Tianhao Yan, Kun Qian, Emilia Parada-Cabaleiro, Bin Hu, Bjoern W. Schuller
Summary: This paper presents two effective neural network models to detect surgical masks from audio, which can extract more salient temporal information. By exploring the combination of LSTM and Transformers in three hybrid models, it is demonstrated that one of the hybrid models achieves the best performance.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2022)
Article
Computer Science, Artificial Intelligence
Giovanni Costantini, Valerio Cesarini, Carlo Robotti, Marco Benazzo, Filomena Pietrantonio, Stefano Di Girolamo, Antonio Pisani, Pietro Canzi, Simone Mauramati, Giulia Bertino, Irene Cassaniti, Fausto Baldanti, Giovanni Saggio
Summary: A novel approach for COVID-19 assessment is adopted in this study, using different vocal tasks and two custom algorithms to identify COVID-19 positive and negative subjects as well as recovered individuals. The results suggest that this method could serve as an on-site screening tool, highlighting the ongoing relevance of traditional machine learning and deep learning in speech analysis.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Federica Amato, Giovanni Saggio, Valerio Cesarini, Gabriella Olmo, Giovanni Costantini
Summary: The preliminary diagnosis and evaluation of Parkinson's disease is crucial. Real-time, non-invasive voice analysis enhanced by machine learning is gaining interest. This review aims to identify the most widely used feature-based machine learning methods and present their effectiveness. A total of 102 works and 5 review articles were selected, analyzing commonly used features, algorithms, datasets, and metadata. Jitter, Shimmer, Harmonic-to-Noise Ratio, Fundamental Frequency, and Mel Frequency Cepstral Coefficients were found to be the most adopted features, with a prevalence of glottal-like models and additional filtering options.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Chemistry, Analytical
Giovanni Costantini, Valerio Cesarini, Emanuele Brenna
Summary: In this paper, the two different methodologies of deep learning and traditional machine learning are compared in speaker recognition task using the DEMoS dataset. The results show that a custom CNN trained on grayscale spectrogram images achieves the most accurate results, with an accuracy of 90.15% for grayscale spectrograms and 83.17% for colored MFCC.
Article
Chemistry, Analytical
Giovanni Costantini, Valerio Cesarini, Pietro Di Leo, Federica Amato, Antonio Suppa, Francesco Asci, Antonio Pisani, Alessandra Calculli, Giovanni Saggio
Summary: This study analyzed the voice characteristics of Parkinson's disease patients using machine learning techniques, and compared different feature selection and classification algorithms. The results showed that both feature-based machine learning and deep learning achieved comparable results in terms of classification, with KNN, SVM, and naive Bayes classifiers performing similarly. The superiority of CFS as the best feature selector was more evident, and the selected features acted as relevant vocal biomarkers capable of differentiating healthy subjects, early untreated PD patients, and mid-advanced L-Dopa treated patients.
Proceedings Paper
Computer Science, Cybernetics
Markus Schedl, Stefan Brandl, Oleg Lesota, Emilia Parada-Cabaleiro, David Penz, Navid Rekabsaz
Summary: The LFM-2b dataset contains the listening records of over 120,000 users on Last.fm, spanning 15 years and involving 50 million distinct tracks and 5 million distinct artists. In addition to common metadata, the dataset also includes demographic information of users and fine-grained genre, style, and lyrics embeddings of items. This rich dataset enables research on various recommender system algorithms and investigation of fairness aspects.
CHIIR'22: PROCEEDINGS OF THE 2022 CONFERENCE ON HUMAN INFORMATION INTERACTION AND RETRIEVAL
(2022)
Proceedings Paper
Engineering, Biomedical
Federica Amato, Maria Fasani, Glauco Raffaelli, Valerio Cesarini, Gabriella Olmo, Nicola Di Lorenzo, Giovanni Costantini, Giovanni Saggio
Summary: Automatic assessment of the influence of obesity and GERD on voice and their mutual influence was conducted using vocal tests from 92 subjects. Machine Learning models achieved high accuracies in scoring the presence of GERD and obesity. Sentence repetition was found to be more effective than vowel phonation, and certain features such as Mel Frequency Cepstral Coefficients were identified as significant for this application.
2022 IEEE INTERNATIONAL SYMPOSIUM ON MEDICAL MEASUREMENTS AND APPLICATIONS (MEMEA 2022)
(2022)
Proceedings Paper
Computer Science, Interdisciplinary Applications
Valerio Cesarini, Carlo Robotti, Ylenia Piromalli, Francesco Mozzanica, Antonio Schindler, Giovanni Saggio, Giovanni Costantini
Summary: This study developed a machine-learning framework to automatically identify and differentiate dysphonic voices. The framework achieved high accuracy and differentiation rates, suggesting the potential for distinguishing the etiologies of dysphonia. The analysis also highlighted a trend of poor volume control in dysphonic subjects, refining existing literature.
BIOSIGNALS: PROCEEDINGS OF THE 15TH INTERNATIONAL JOINT CONFERENCE ON BIOMEDICAL ENGINEERING SYSTEMS AND TECHNOLOGIES - VOL 4: BIOSIGNALS
(2022)