We presented the Treff adapter, a training-efficient adapter for CLAP, to boost zero-shot classification performance by making use of a small set of labelled data. Specifically, we designed CALM to retrieve the probability distribution of text-audio clips over classes using a set of audio-label pairs and combined it with CLAP's zero-shot classification results. Furthermore, we designed a training-free version of the Treff adapter by using CALM as a cosine similarity measure. Experiments showed that the proposed Treff adapter is comparable and even better than fully-supervised methods and adaptation methods in low-shot and data-abundant scenarios. While the Treff adapter shows that combining large-scale pretraining and rapid learning of domain-specific knowledge is non-trivial for obtaining generic representations for few-shot learning, it is still limited to audio classification tasks. In the future, we will explore how to use audio-language models in diverse audio domains.
Everyday sound recognition aims to infer types of sound events in audio streams. While many works succeeded in training models with high performance in a fully-supervised manner, they are still restricted to the demand of large quantities of labelled data and the range of predefined classes. To overcome these drawbacks, this work firstly curates a new database named FSD-FS for multi-label few-shot audio classification. It then explores how to incorporate audio taxonomy in few-shot learning. Specifically, this work proposes label-dependent prototypical networks (LaD-protonet) to exploit parent-children relationships between labels. Plus, it applies taxonomy-aware label smoothing techniques to boost model performance. Experiments demonstrate that LaD-protonet outperforms original prototypical networks as well as other state-of-the-art methods. Moreover, its performance can be further boosted when combined with taxonomy-aware label smoothing.
The deep learning community has witnessed an exponentially growing interest in self-supervised learning (SSL). However, it still remains unexplored how to build a framework for learning useful representations of raw music waveforms in a self-supervised manner. In this work, we design Music2Vec, a framework exploring different SSL algorithmic components and tricks for music audio recordings. Our model achieves comparable results to the state-of-the-art (SOTA) music SSL model Jukebox, despite being significantly smaller with less than 2% of parameters of the latter. The model will be released on Huggingface(Please refer to: https://huggingface.co/m-a-p/music2vec-v1)
Learning music representations that are general-purpose offers the flexibility to finetune several downstream tasks using smaller datasets. The wav2vec 2.0 speech representation model showed promising results in many downstream speech tasks, but has been less effective when adapted to music. In this paper, we evaluate whether pre-training wav2vec 2.0 directly on music data can be a better solution instead of finetuning the speech model. We illustrate that when pre-training on music data, the discrete latent representations are able to encode the semantic meaning of musical concepts such as pitch and instrument. Our results show that finetuning wav2vec 2.0 pre-trained on music data allows us to achieve promising results on music classification tasks that are competitive with prior work on audio representations. In addition, the results are superior to the pre-trained model on speech embeddings, demonstrating that wav2vec 2.0 pre-trained on music data can be a promising music representation model.
As one of the most intuitive interfaces known to humans, natural language has the potential to mediate many tasks that involve human-computer interaction, especially in application-focused fields like Music Information Retrieval. In this work, we explore cross-modal learning in an attempt to bridge audio and language in the music domain. To this end, we propose MusCALL, a framework for Music Contrastive Audio-Language Learning. Our approach consists of a dual-encoder architecture that learns the alignment between pairs of music audio and descriptive sentences, producing multimodal embeddings that can be used for text-to-audio and audio-to-text retrieval out-of-the-box. Thanks to this property, MusCALL can be transferred to virtually any task that can be cast as text-based retrieval. Our experiments show that our method performs significantly better than the baselines at retrieving audio that matches a textual description and, conversely, text that matches an audio query. We also demonstrate that the multimodal alignment capability of our model can be successfully extended to the zero-shot transfer scenario for genre classification and auto-tagging on two public datasets.
Loss-gradients are used to interpret the decision making process of deep learning models. In this work, we evaluate loss-gradient based attribution methods by occluding parts of the input and comparing the performance of the occluded input to the original input. We observe that the occluded input has better performance than the original across the test dataset under certain conditions. Similar behaviour is observed in sound and image recognition tasks. We explore different loss-gradient attribution methods, occlusion levels and replacement values to explain the phenomenon of performance improvement under occlusion.
Imitating musical instruments with the human voice is an efficient way of communicating ideas between music producers, from sketching melody lines to clarifying desired sonorities. For this reason, there is an increasing interest in building applications that allow artists to efficiently pick target samples from big sound libraries just by imitating them vocally. In this study, we investigated the potential of conditional autoencoder models to learn informative features for Drum Sample Retrieval by Vocalisation (DSRV). We assessed the usefulness of their embeddings using four evaluation metrics, two of them relative to their acoustic properties and two of them relative to their perceptual properties via human listeners' similarity ratings. Results suggest that models conditioned on both sound-type labels (drum vs imitation) and drum-type labels (kick vs snare vs closed hi-hat vs opened hi-hat) learn the most informative embeddings for DSRV. We finally looked into individual differences in vocal imitation style via the Mantel test and found salient differences among participants, highlighting the importance of user information when designing DSRV systems.
Most recent research about automatic music transcription (AMT) uses convolutional neural networks and recurrent neural networks to model the mapping from music signals to symbolic notation. Based on a high-resolution piano transcription system, we explore the possibility of incorporating another powerful sequence transformation tool -- the Transformer -- to deal with the AMT problem. We argue that the properties of the Transformer make it more suitable for certain AMT subtasks. We confirm the Transformer's superiority on the velocity detection task by experiments on the MAESTRO dataset and a cross-dataset evaluation on the MAPS dataset. We observe a performance improvement on both frame-level and note-level metrics after introducing the Transformer network.
This paper introduces a comparison of deep learning-based techniques for the MOS prediction task of synthesised speech in the Interspeech VoiceMOS challenge. Using the data from the main track of the VoiceMOS challenge we explore both existing predictors and propose new ones. We evaluate two groups of models: NISQA-based models and techniques based on fine-tuning the self-supervised learning (SSL) model wav2vec2_base. Our findings show that a simplified version of NISQA with 40% fewer parameters achieves results close to the original NISQA architecture on both utterance-level and system-level performances. Pre-training NISQA with the NISQA corpus improves utterance-level performance but shows no benefit on the system-level performance. Also, the NISQA-based models perform close to LDNet and MOSANet, 2 out of 3 baselines of the challenge. Fine-tuning wav2vec2_base shows superior performance than the NISQA-based models. We explore the mismatch between natural and synthetic speech and discovered that the performance of the SSL model drops consistently when fine-tuned on natural speech samples. We show that adding CNN features with the SSL model does not improve the baseline performance. Finally, we show that the system type has an impact on the predictions of the non-SSL models.
In recent years, the accuracy of automatic lyrics alignment methods has increased considerably. Yet, many current approaches employ frameworks designed for automatic speech recognition (ASR) and do not exploit properties specific to music. Pitch is one important musical attribute of singing voice but it is often ignored by current systems as the lyrics content is considered independent of the pitch. In practice, however, there is a temporal correlation between the two as note starts often correlate with phoneme starts. At the same time the pitch is usually annotated with high temporal accuracy in ground truth data while the timing of lyrics is often only available at the line (or word) level. In this paper, we propose a multi-task learning approach for lyrics alignment that incorporates pitch and thus can make use of a new source of highly accurate temporal information. Our results show that the accuracy of the alignment result is indeed improved by our approach. As an additional contribution, we show that integrating boundary detection in the forced-alignment algorithm reduces cross-line errors, which improves the accuracy even further.