Christopher J. Tralie

Christopher J. Tralie, Matt Amery, Benjamin Douglas et al.

As generative techniques pervade the audio domain, there has been increasing interest in tracing back through these complicated models to understand how they draw on their training data to synthesize new examples, both to ensure that they use properly licensed data and also to elucidate their black box behavior. In this paper, we show that if imperceptible echoes are hidden in the training data, a wide variety of audio to audio architectures (differentiable digital signal processing (DDSP), Realtime Audio Variational autoEncoder (RAVE), and ``Dance Diffusion'') will reproduce these echoes in their outputs. Hiding a single echo is particularly robust across all architectures, but we also show promising results hiding longer time spread echo patterns for an increased information capacity. We conclude by showing that echoes make their way into fine tuned models, that they survive mixing/demixing, and that they survive pitch shift augmentation during training. Hence, this simple, classical idea in watermarking shows significant promise for tagging generative audio models.

Furkan Yesiler, Guillaume Doras, Rachel M. Bittner et al.

In this article, we aim to provide a review of the key ideas and approaches proposed in 20 years of scientific literature around musical version identification (VI) research and connect them to current practice. For more than a decade, VI systems suffered from the accuracy-scalability trade-off, with attempts to increase accuracy that typically resulted in cumbersome, non-scalable systems. Recent years, however, have witnessed the rise of deep learning-based approaches that take a step toward bridging the accuracy-scalability gap, yielding systems that can realistically be deployed in industrial applications. Although this trend positively influences the number of researchers and institutions working on VI, it may also result in obscuring the literature before the deep learning era. To appreciate two decades of novel ideas in VI research and to facilitate building better systems, we now review some of the successful concepts and applications proposed in the literature and study their evolution throughout the years.

Christopher J. Tralie, Brian McFee

We describe a novel pipeline to automatically discover hierarchies of repeated sections in musical audio. The proposed method uses similarity network fusion (SNF) to combine different frame-level features into clean affinity matrices, which are then used as input to spectral clustering. While prior spectral clustering approaches to music structure analysis have pre-processed affinity matrices with heuristics specifically designed for this task, we show that the SNF approach directly yields segmentations which agree better with human annotators, as measured by the ``L-measure'' metric for hierarchical annotations. Furthermore, the SNF approach immediately supports arbitrarily many input features, allowing us to simultaneously discover structure encoded in timbral, harmonic, and rhythmic representations without any changes to the base algorithm.

Christopher J. Tralie, Paul Bendich, John Harer

In this work, we address fusion of heterogeneous sensor data using wavelet-based summaries of fused self-similarity information from each sensor. The technique we develop is quite general, does not require domain specific knowledge or physical models, and requires no training. Nonetheless, it can perform surprisingly well at the general task of differentiating classes of time-ordered behavior sequences which are sensed by more than one modality. As a demonstration of our capabilities in the audio to video context, we focus on the differentiation of speech sequences. Data from two or more modalities first are represented using self-similarity matrices(SSMs) corresponding to time-ordered point clouds in feature spaces of each of these data sources; we note that these feature spaces can be of entirely different scale and dimensionality. A fused similarity template is then derived from the modality-specific SSMs using a technique called similarity network fusion (SNF). We investigate pipelines using SNF as both an upstream (feature-level) and a downstream (ranking-level) fusion technique. Multiscale geometric features of this template are then extracted using a recently-developed technique called the scattering transform, and these features are then used to differentiate speech sequences. This method outperforms unsupervised techniques which operate directly on the raw data, and it also outperforms stovepiped methods which operate on SSMs separately derived from the distinct modalities. The benefits of this method become even more apparent as the simulated peak signal to noise ratio decreases.

Christopher J. Tralie

In this work, we pose and address the following "cover song analogies" problem: given a song A by artist 1 and a cover song A' of this song by artist 2, and given a different song B by artist 1, synthesize a song B' which is a cover of B in the style of artist 2. Normally, such a polyphonic style transfer problem would be quite challenging, but we show how the cover songs example constrains the problem, making it easier to solve. First, we extract the longest common beat-synchronous subsequence between A and A', and we time stretch the corresponding beat intervals in A' so that they align with A. We then derive a version of joint 2D convolutional NMF, which we apply to the constant-Q spectrograms of the synchronized segments to learn a translation dictionary of sound templates from A to A'. Finally, we apply the learned templates as filters to the song B, and we mash up the translated filtered components into the synthesized song B' using audio mosaicing. We showcase our algorithm on several examples, including a synthesized cover version of Michael Jackson's "Bad" by Alien Ant Farm, learned from the latter's "Smooth Criminal" cover.'

Christopher J. Tralie, Abraham Smith, Nathan Borggren et al.

In this work, we address the problem of cross-modal comparison of aerial data streams. A variety of simulated automobile trajectories are sensed using two different modalities: full-motion video, and radio-frequency (RF) signals received by detectors at various locations. The information represented by the two modalities is compared using self-similarity matrices (SSMs) corresponding to time-ordered point clouds in feature spaces of each of these data sources; we note that these feature spaces can be of entirely different scale and dimensionality. Several metrics for comparing SSMs are explored, including a cutting-edge time-warping technique that can simultaneously handle local time warping and partial matches, while also controlling for the change in geometry between feature spaces of the two modalities. We note that this technique is quite general, and does not depend on the choice of modalities. In this particular setting, we demonstrate that the cross-modal distance between SSMs corresponding to the same trajectory type is smaller than the cross-modal distance between SSMs corresponding to distinct trajectory types, and we formalize this observation via precision-recall metrics in experiments. Finally, we comment on promising implications of these ideas for future integration into multiple-hypothesis tracking systems.

Christopher J. Tralie

In this work, we explore the problem of aligning two time-ordered point clouds which are spatially transformed and re-parameterized versions of each other. This has a diverse array of applications such as cross modal time series synchronization (e.g. MOCAP to video) and alignment of discretized curves in images. Most other works that address this problem attempt to jointly uncover a spatial alignment and correspondences between the two point clouds, or to derive local invariants to spatial transformations such as curvature before computing correspondences. By contrast, we sidestep spatial alignment completely by using self-similarity matrices (SSMs) as a proxy to the time-ordered point clouds, since self-similarity matrices are blind to isometries and respect global geometry. Our algorithm, dubbed "Isometry Blind Dynamic Time Warping" (IBDTW), is simple and general, and we show that its associated dissimilarity measure lower bounds the L1 Gromov-Hausdorff distance between the two point sets when restricted to warping paths. We also present a local, partial alignment extension of IBDTW based on the Smith Waterman algorithm. This eliminates the need for tedious manual cropping of time series, which is ordinarily necessary for global alignment algorithms to function properly.

Christopher J. Tralie

While most schemes for automatic cover song identification have focused on note-based features such as HPCP and chord profiles, a few recent papers surprisingly showed that local self-similarities of MFCC-based features also have classification power for this task. Since MFCC and HPCP capture complementary information, we design an unsupervised algorithm that combines normalized, beat-synchronous blocks of these features using cross-similarity fusion before attempting to locally align a pair of songs. As an added bonus, our scheme naturally incorporates structural information in each song to fill in alignment gaps where both feature sets fail. We show a striking jump in performance over MFCC and HPCP alone, achieving a state of the art mean reciprocal rank of 0.87 on the Covers80 dataset. We also introduce a new medium-sized hand designed benchmark dataset called "Covers 1000," which consists of 395 cliques of cover songs for a total of 1000 songs, and we show that our algorithm achieves an MRR of 0.9 on this dataset for the first correctly identified song in a clique. We provide the precomputed HPCP and MFCC features, as well as beat intervals, for all songs in the Covers 1000 dataset for use in further research.

Christopher J. Tralie, Jose A. Perea

This work introduces a novel framework for quantifying the presence and strength of recurrent dynamics in video data. Specifically, we provide continuous measures of periodicity (perfect repetition) and quasiperiodicity (superposition of periodic modes with non-commensurate periods), in a way which does not require segmentation, training, object tracking or 1-dimensional surrogate signals. Our methodology operates directly on video data. The approach combines ideas from nonlinear time series analysis (delay embeddings) and computational topology (persistent homology), by translating the problem of finding recurrent dynamics in video data, into the problem of determining the circularity or toroidality of an associated geometric space. Through extensive testing, we show the robustness of our scores with respect to several noise models/levels, we show that our periodicity score is superior to other methods when compared to human-generated periodicity rankings, and furthermore, we show that our quasiperiodicity score clearly indicates the presence of biphonation in videos of vibrating vocal folds, which has never before been accomplished end to end quantitatively.

Christopher J. Tralie, Paul Bendich

We introduce a novel low level feature for identifying cover songs which quantifies the relative changes in the smoothed frequency spectrum of a song. Our key insight is that a sliding window representation of a chunk of audio can be viewed as a time-ordered point cloud in high dimensions. For corresponding chunks of audio between different versions of the same song, these point clouds are approximately rotated, translated, and scaled copies of each other. If we treat MFCC embeddings as point clouds and cast the problem as a relative shape sequence, we are able to correctly identify 42/80 cover songs in the "Covers 80" dataset. By contrast, all other work to date on cover songs exclusively relies on matching note sequences from Chroma derived features.