Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · 2208.14750

Harmonization and Evaluation; Tweaking the Parameters on Human Listeners

Filippo Carnovalini·Alessandro Pelizzo·Antonio Rodà·Sergio Canazza
View original source ↗Read source document ↗Export metadata ↓

Abstract

Kansei models were used to study the connotative meaning of music. In multimedia and mixed reality, automatically generated melodies are increasingly being used. It is important to consider whether and what feelings are communicated by this music. Evaluation of computer-generated melodies is not a trivial task. Considered the difficulty of defining useful quantitative metrics of the quality of a generated musical piece, researchers often resort to human evaluation. In these evaluations, often the judges are required to evaluate a set of generated pieces along with some benchmark pieces. The latter are often composed by humans. While this kind of evaluation is relatively common, it is known that care should be taken when designing the experiment, as humans can be influenced by a variety of factors. In this paper, we examine the impact of the presence of harmony in audio files that judges must evaluate, to see whether having an accompaniment can change the evaluation of generated melodies. To do so, we generate melodies with two different algorithms and harmonize them with an automatic tool that we designed for this experiment, and ask more than sixty participants to evaluate the melodies. By using statistical analyses, we show harmonization does impact the evaluation process, by emphasizing the differences among judgements.

Explore related subjects

cs.SDcs.MMeess.AS

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Filippo Carnovalini, Alessandro Pelizzo, Antonio Rodà, Sergio Canazza. 2022-08-31. Harmonization and Evaluation; Tweaking the Parameters on Human Listeners. https://arxiv.org/abs/2208.14750

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Xiaomi-CocktailASR-1 Technical Report

Recently, large language model (LLM) based ASR models have achieved significant progress, yet they generally lack support for multi-speaker scenarios, where the cocktail party problem remains a critical bottleneck for further advancing ASR. Existing TS-ASR methods, including end-to-end architectures with speaker embeddings and latest LLM-based explorations suffer from degraded single-speaker performance and the inability to reject when the target speaker is absent. In this paper, we propose Xiaomi-CocktailASR-1, an LLM-based end-to-end TS-ASR architecture. By utilizing reference speech as voiceprint prompts, it directly transcribes the target speaker's speech without requiring speech separation. Xiaomi-CocktailASR-1 maintains competitive performance in single-speaker scenarios, comparable to mainstream ASR models. It also features a negative sample rejection capability, outputting empty text when the target speaker is absent from the mixed speech. Additionally, Xiaomi-CocktailASR-1 supports a Chain-of-Thought (CoT) reasoning mode to provide explicit reasoning steps. Extensive experiments on various synthetic and real-world multispeaker benchmarks demonstrate that Xiaomi-CocktailASR-1 achieves state-of-the-art performance, effectively addressing the cocktail party problem through a unified architecture that balances multispeaker and single-speaker recognition accuracy, along with rejection capability.

cs.SD↗
arXiv2026

EConv-TasNet: Efficient Conv-TasNet for Effective Speech Separation

Conv-TasNet has served as a strong baseline for time-domain speech separation, and many studies have extended it with advanced architectures such as dual-path networks, U-Nets, and attention mechanisms. However, these methods often introduce high computational cost and complexity, limiting their deployment in resource-constrained scenarios. To address this issue, we propose eConv-TasNet, an efficient variant of Conv-TasNet that improves both effectiveness and efficiency without relying on resource-intensive modules. The proposed model consists of a group-wise early-splitting (GES) module and a multi-group feature aggregation (MGFA) module. GES generates discriminative speaker embeddings at intermediate stages, while MGFA progressively aggregates these group-level representations for refined mask estimation. Experimental results show that eConv-TasNet reduces model size by 22.4%, accelerates inference by 18.9%, and improves SI-SNRi by 14.0%-28.0% across three public benchmarks. Moreover, it achieves competitive performance compared with state-of-the-art methods while requiring significantly fewer parameters and lower inference cost. These results demonstrate a favorable efficiency-effectiveness trade-off for edge deployment.

cs.SD↗