Konstantinos Velenis
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Konstantinos Velenis has a total of 4 co-authors. Here are the top co-authors by number of shared publications:
- Emilios Cambouropoulos
Department of Music Studies, School of Fine Arts, Aristotle University of Thessaloniki , Thessaloniki
3 shared publications - Asterios Zacharakis
School of Music Studies, Aristotle University of Thessaloniki
2 shared publications - Maximos Kaliakatsos-Papakostas
Department of Music Technology and Acoustics, Hellenic Mediterranean University, 74133 Rethymno, Greece
2 shared publications - Leandros Pasias
School of Music Studies, Aristotle University, Greece
1 shared publication
Latest Publications
- Open Access
- 1 April 2026
- Published by Acoustical Society of America (ASA) in JASA Express Letters
- Vol. 6 (4)
Beyond fundamental frequency: Pitch height judgments in Indefinite-Pitch sounds
Pitch judgments for sounds lacking a harmonic spectrum and a clear fundamental frequency are inherently ambiguous. Using a sinusoidal adjustment task, 48 listeners provided pitch-height judgments for 19 non-harmonic sounds. Linear mixed-effects analyses showed that listeners could reliably differentiate pitch height across stimuli despite weak or absent periodic cues. A loudness-weighted spectral centroid emerged as a significant predictor of perceived pitch height, with reduced accuracy for sounds exhibiting greater spectrotemporal variability. These results indicate that perceptually weighted spectral information supports pitch-height judgments in non-harmonic sounds.
- 1 April 2026
- Published by Acoustical Society of America (ASA) in JASA Express Letters
- Vol. 6 (4)
Open AccessBeyond fundamental frequency: Pitch height judgments in Indefinite-Pitch sounds
Pitch judgments for sounds lacking a harmonic spectrum and a clear fundamental frequency are inherently ambiguous. Using a sinusoidal adjustment task, 48 listeners provided pitch-height judgments for 19 non-harmonic sounds. Linear mixed-effects analyses showed that listeners could reliably differentiate pitch height across stimuli despite weak or absent periodic cues. A loudness-weighted spectral centroid emerged as a significant predictor of perceived pitch height, with reduced accuracy for sounds exhibiting greater spectrotemporal variability. These results indicate that perceptually weighted spectral information supports pitch-height judgments in non-harmonic sounds.
- Open Access
- 18 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 285-291
Rebetiko Singer Identification: Fine-tuning and explaining deep pretrained transformer models
In Rebetiko music, the distinctive vocal qualities of individual singers, especially renowned singers, are often at the centre of attention, as their unique timbre and ornamentations play a crucial role in defining the genre’s characteristics. Pre-trained deep learning models on human voice data offer a good starting point for fine-tuning to classify small-sized voice-related datasets. While these fine-tuned models are "black boxes" that do not provide direct explanations for their classifications, their internal latent representations can be explored through visualizations. This study analyzes the performance and possibilities for explainability of pre-trained voice models fine-tuned for identifying Rebetiko singers using separated voice signals from original recordings. The selected singers exhibit diverse timbral and delivery styles, which can be numerically represented through perceptually relevant audio features (e.g., fundamental frequency, spectral centroid, vibrato rate and depth, inharmonicity, etc.). By comparing extracted audio features with visualizations from the "black box" model, valuable insights into the model’s learning process are gained. This study not only demonstrates the effectiveness of fine-tuning for Rebetiko singer identification but also enhances explainability, which is crucial for better understanding the characteristics and trends of Rebetiko singing.
- 18 September 2024
- Published by Association for Computing Machinery (ACM)
- p. 285-291
Open AccessRebetiko Singer Identification: Fine-tuning and explaining deep pretrained transformer models
In Rebetiko music, the distinctive vocal qualities of individual singers, especially renowned singers, are often at the centre of attention, as their unique timbre and ornamentations play a crucial role in defining the genre’s characteristics. Pre-trained deep learning models on human voice data offer a good starting point for fine-tuning to classify small-sized voice-related datasets. While these fine-tuned models are "black boxes" that do not provide direct explanations for their classifications, their internal latent representations can be explored through visualizations. This study analyzes the performance and possibilities for explainability of pre-trained voice models fine-tuned for identifying Rebetiko singers using separated voice signals from original recordings. The selected singers exhibit diverse timbral and delivery styles, which can be numerically represented through perceptually relevant audio features (e.g., fundamental frequency, spectral centroid, vibrato rate and depth, inharmonicity, etc.). By comparing extracted audio features with visualizations from the "black box" model, valuable insights into the model’s learning process are gained. This study not only demonstrates the effectiveness of fine-tuning for Rebetiko singer identification but also enhances explainability, which is crucial for better understanding the characteristics and trends of Rebetiko singing.
- Open Access
- 24 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 79-87
Visualisations of Jazz Standards Derived from Transformer-based Multi-task Learning
This paper presents a method for creating multiple 2D visualizations of 1048 jazz standard chord charts in text format. The novel component of this method is that each visualization re-arranges the data available, prioritizing specific musical aspects, namely harmonic context, genre style, composition year, music form, composer, and tonality, allowing the exploration of this dataset according to user-defined criteria. The basic component of this model is a stack of transformer encoders that generate contextualized information from chart-text input. This information is subsequently fed to downstream feed-forward models that perform classification and regression based on manually annotated target labels and values of the aforementioned visualization criteria. The training of our model was performed on a distinct dataset from the validation and testing sets, ensuring robustness and minimizing the risk of overfitting. A model variation where the transformer component performs jointly unmasking of the text input outperforms a variation without umnasking, when both models are trained on all 11 transpositions of the available pieces and tested on the original tonalities. The resulting visualisations reflect the categorisation and regression capabilities of the outperforming model, while revealing some interesting inter-cluster relations as well as some justifiable out-of-place pieces. A few hard to explain placements of specific pieces provide pointers for future work. Overall, such visualisations show the potential to empower jazz musicians to explore jazz standards under multiple perspectives.
- 24 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 79-87
Open AccessVisualisations of Jazz Standards Derived from Transformer-based Multi-task Learning
This paper presents a method for creating multiple 2D visualizations of 1048 jazz standard chord charts in text format. The novel component of this method is that each visualization re-arranges the data available, prioritizing specific musical aspects, namely harmonic context, genre style, composition year, music form, composer, and tonality, allowing the exploration of this dataset according to user-defined criteria. The basic component of this model is a stack of transformer encoders that generate contextualized information from chart-text input. This information is subsequently fed to downstream feed-forward models that perform classification and regression based on manually annotated target labels and values of the aforementioned visualization criteria. The training of our model was performed on a distinct dataset from the validation and testing sets, ensuring robustness and minimizing the risk of overfitting. A model variation where the transformer component performs jointly unmasking of the text input outperforms a variation without umnasking, when both models are trained on all 11 transpositions of the available pieces and tested on the original tonalities. The resulting visualisations reflect the categorisation and regression capabilities of the outperforming model, while revealing some interesting inter-cluster relations as well as some justifiable out-of-place pieces. A few hard to explain placements of specific pieces provide pointers for future work. Overall, such visualisations show the potential to empower jazz musicians to explore jazz standards under multiple perspectives.