Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2509.04985 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866908521197469696 |
|---|---|
| author | Liu, Yuxuan Sang, Rui Zhang, Peihong Li, Zhixin Li, Shengchen |
| author_facet | Liu, Yuxuan Sang, Rui Zhang, Peihong Li, Zhixin Li, Shengchen |
| contents | Music Information Retrieval (MIR) systems are highly vulnerable to adversarial attacks that are often imperceptible to humans, primarily due to a misalignment between model feature spaces and human auditory perception. Existing defenses and perceptual metrics frequently fail to adequately capture these auditory nuances, a limitation supported by our initial listening tests showing low correlation between common metrics and human judgments. To bridge this gap, we introduce Perceptually-Aligned MERT Transformer (PAMT), a novel framework for learning robust, perceptually-aligned music representations. Our core innovation lies in the psychoacoustically-conditioned sequential contrastive transformer, a lightweight projection head built atop a frozen MERT encoder. PAMT achieves a Spearman correlation coefficient of 0.65 with subjective scores, outperforming existing perceptual metrics. Our approach also achieves an average of 9.15\% improvement in robust accuracy on challenging MIR tasks, including Cover Song Identification and Music Genre Classification, under diverse perceptual adversarial attacks. This work pioneers architecturally-integrated psychoacoustic conditioning, yielding representations significantly more aligned with human perception and robust against music adversarial attacks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_04985 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Training a Perceptual Model for Evaluating Auditory Similarity in Music Adversarial Attack Liu, Yuxuan Sang, Rui Zhang, Peihong Li, Zhixin Li, Shengchen Sound Audio and Speech Processing Music Information Retrieval (MIR) systems are highly vulnerable to adversarial attacks that are often imperceptible to humans, primarily due to a misalignment between model feature spaces and human auditory perception. Existing defenses and perceptual metrics frequently fail to adequately capture these auditory nuances, a limitation supported by our initial listening tests showing low correlation between common metrics and human judgments. To bridge this gap, we introduce Perceptually-Aligned MERT Transformer (PAMT), a novel framework for learning robust, perceptually-aligned music representations. Our core innovation lies in the psychoacoustically-conditioned sequential contrastive transformer, a lightweight projection head built atop a frozen MERT encoder. PAMT achieves a Spearman correlation coefficient of 0.65 with subjective scores, outperforming existing perceptual metrics. Our approach also achieves an average of 9.15\% improvement in robust accuracy on challenging MIR tasks, including Cover Song Identification and Music Genre Classification, under diverse perceptual adversarial attacks. This work pioneers architecturally-integrated psychoacoustic conditioning, yielding representations significantly more aligned with human perception and robust against music adversarial attacks. |
| title | Training a Perceptual Model for Evaluating Auditory Similarity in Music Adversarial Attack |
| topic | Sound Audio and Speech Processing |
| url | https://arxiv.org/abs/2509.04985 |