Kazuyoshi Yoshii

Kouhei Sekiguchi, Aditya Arie Nugraha, Yicheng Du et al.

This paper describes the practical response- and performance-aware development of online speech enhancement for an augmented reality (AR) headset that helps a user understand conversations made in real noisy echoic environments (e.g., cocktail party). One may use a state-of-the-art blind source separation method called fast multichannel nonnegative matrix factorization (FastMNMF) that works well in various environments thanks to its unsupervised nature. Its heavy computational cost, however, prevents its application to real-time processing. In contrast, a supervised beamforming method that uses a deep neural network (DNN) for estimating spatial information of speech and noise readily fits real-time processing, but suffers from drastic performance degradation in mismatched conditions. Given such complementary characteristics, we propose a dual-process robust online speech enhancement method based on DNN-based beamforming with FastMNMF-guided adaptation. FastMNMF (back end) is performed in a mini-batch style and the noisy and enhanced speech pairs are used together with the original parallel training data for updating the direction-aware DNN (front end) with backpropagation at a computationally-allowable interval. This method is used with a blind dereverberation method called weighted prediction error (WPE) for transcribing the noisy reverberant speech of a speaker, which can be detected from video or selected by a user's hand gesture or eye gaze, in a streaming manner and spatially showing the transcriptions with an AR technique. Our experiment showed that the word error rate was improved by more than 10 points with the run-time adaptation using only twelve minutes of observation.

Yoshiaki Bando, Yoshiki Masuyama, Aditya Arie Nugraha et al.

This paper describes an efficient unsupervised learning method for a neural source separation model that utilizes a probabilistic generative model of observed multichannel mixtures proposed for blind source separation (BSS). For this purpose, amortized variational inference (AVI) has been used for directly solving the inverse problem of BSS with full-rank spatial covariance analysis (FCA). Although this unsupervised technique called neural FCA is in principle free from the domain mismatch problem, it is computationally demanding due to the full rankness of the spatial model in exchange for robustness against relatively short reverberations. To reduce the model complexity without sacrificing performance, we propose neural FastFCA based on the jointly-diagonalizable yet full-rank spatial model. Our neural separation model introduced for AVI alternately performs neural network blocks and single steps of an efficient iterative algorithm called iterative source steering. This alternating architecture enables the separation model to quickly separate the mixture spectrogram by leveraging both the deep neural network and the multichannel optimization algorithm. The training objective with AVI is derived to maximize the marginalized likelihood of the observed mixtures. The experiment using mixture signals of two to four sound sources shows that neural FastFCA outperforms conventional BSS methods and reduces the computational time to about 2% of that for the neural FCA.

Yicheng Du, Aditya Arie Nugraha, Kouhei Sekiguchi et al.

This paper describes noisy speech recognition for an augmented reality headset that helps verbal communication within real multiparty conversational environments. A major approach that has actively been studied in simulated environments is to sequentially perform speech enhancement and automatic speech recognition (ASR) based on deep neural networks (DNNs) trained in a supervised manner. In our task, however, such a pretrained system fails to work due to the mismatch between the training and test conditions and the head movements of the user. To enhance only the utterances of a target speaker, we use beamforming based on a DNN-based speech mask estimator that can adaptively extract the speech components corresponding to a head-relative particular direction. We propose a semi-supervised adaptation method that jointly updates the mask estimator and the ASR model at run-time using clean speech signals with ground-truth transcriptions and noisy speech signals with highly-confident estimated transcriptions. Comparative experiments using the state-of-the-art distant speech recognition system show that the proposed method significantly improves the ASR performance.

Weixing Wei, Raynaldi Lalang, Dichucheng Li et al.

This paper describes a novel paradigm that formalizes automatic piano transcription (APT) as an optimal transport (OT) problem, not as a frame-level multi-label binary classification problem. Our method learns to minimize the cost of transporting a predicted distribution of note events to the ground-truth distribution over time and frequency. The OT loss can thus accommodate temporal misalignment, leading to perceptually relevant optimization. We also propose a convolutional recurrent neural network (CRNN) with a harmonics-aware attention mechanism to capture the spectro-temporal dependencies inherent in music.Our experiments using the MAESTRO dataset showed that our method attained a state-of-the-art performance in onset detection. We confirmed the versatility of the OT loss in application to existing models.

Jiahao Zhao, Yunjia Li, Wei Li et al.

As large language models continue to develop, the feasibility and significance of text-based symbolic music tasks have become increasingly prominent. While symbolic music has been widely used in generation tasks, LLM capabilities in understanding and reasoning about symbolic music remain largely underexplored. To address this gap, we propose ABC-Eval, the first open-source benchmark dedicated to the understanding and instruction-following capabilities in text-based ABC notation scores. It comprises 1,086 test samples spanning 10 sub-tasks, covering scenarios from basic musical syntax comprehension to complex sequence-level reasoning. Such a diverse scope poses substantial challenges to models' ability to handle symbolic music tasks. We evaluated seven state-of-the-art LLMs on ABC-Eval, and the results reveal notable limitations in existing models' symbolic music processing capabilities. Furthermore, the consistent performance of individual baselines across different sub-tasks supports the reliability of our benchmark.

Yoto Fujita, Yoshiaki Bando, Keisuke Imoto et al.

This paper describes sound event localization and detection (SELD) for spatial audio recordings captured by firstorder ambisonics (FOA) microphones. In this task, one may train a deep neural network (DNN) using FOA data annotated with the classes and directions of arrival (DOAs) of sound events. However, the performance of this approach is severely bounded by the amount of annotated data. To overcome this limitation, we propose a novel method of pretraining the feature extraction part of the DNN in a self-supervised manner. We use spatial audio-visual recordings abundantly available as virtual reality contents. Assuming that sound objects are concurrently observed by the FOA microphones and the omni-directional camera, we jointly train audio and visual encoders with contrastive learning such that the audio and visual embeddings of the same recording and DOA are made close. A key feature of our method is that the DOA-wise audio embeddings are jointly extracted from the raw audio data, while the DOA-wise visual embeddings are separately extracted from the local visual crops centered on the corresponding DOA. This encourages the latent features of the audio encoder to represent both the classes and DOAs of sound events. The experiment using the DCASE2022 Task 3 dataset of 20 hours shows non-annotated audio-visual recordings of 100 hours reduced the error score of SELD from 36.4 pts to 34.9 pts.

Yoto Fujita, Aditya Arie Nugraha, Diego Di Carlo et al.

This paper describes speech enhancement for realtime automatic speech recognition (ASR) in real environments. A standard approach to this task is to use neural beamforming that can work efficiently in an online manner. It estimates the masks of clean dry speech from a noisy echoic mixture spectrogram with a deep neural network (DNN) and then computes a enhancement filter used for beamforming. The performance of such a supervised approach, however, is drastically degraded under mismatched conditions. This calls for run-time adaptation of the DNN. Although the ground-truth speech spectrogram required for adaptation is not available at run time, blind dereverberation and separation methods such as weighted prediction error (WPE) and fast multichannel nonnegative matrix factorization (FastMNMF) can be used for generating pseudo groundtruth data from a mixture. Based on this idea, a prior work proposed a dual-process system based on a cascade of WPE and minimum variance distortionless response (MVDR) beamforming asynchronously fine-tuned by block-online FastMNMF. To integrate the dereverberation capability into neural beamforming and make it fine-tunable at run time, we propose to use weighted power minimization distortionless response (WPD) beamforming, a unified version of WPE and minimum power distortionless response (MPDR), whose joint dereverberation and denoising filter is estimated using a DNN. We evaluated the impact of run-time adaptation under various conditions with different numbers of speakers, reverberation times, and signal-to-noise ratios (SNRs).

Diego Di Carlo, Mathieu Fontaine, Aditya Arie Nugraha et al.

This paper describes a sound source localization (SSL) technique that combines an $α$-stable model for the observed signal with a neural network-based approach for modeling steering vectors. Specifically, a physics-informed neural network, referred to as Neural Steerer, is used to interpolate measured steering vectors (SVs) on a fixed microphone array. This allows for a more robust estimation of the so-called $α$-stable spatial measure, which represents the most plausible direction of arrival (DOA) of a target signal. As an $α$-stable model for the non-Gaussian case ($α$ $\in$ (0, 2)) theoretically defines a unique spatial measure, we choose to leverage it to account for residual reconstruction error of the Neural Steerer in the downstream tasks. The objective scores indicate that our proposed technique outperforms state-of-the-art methods in the case of multiple sound sources.

Ryoto Ishizuka, Ryo Nishikimi, Kazuyoshi Yoshii

This paper describes an automatic drum transcription (ADT) method that directly estimates a tatum-level drum score from a music signal, in contrast to most conventional ADT methods that estimate the frame-level onset probabilities of drums. To estimate a tatum-level score, we propose a deep transcription model that consists of a frame-level encoder for extracting the latent features from a music signal and a tatum-level decoder for estimating a drum score from the latent features pooled at the tatum level. To capture the global repetitive structure of drum scores, which is difficult to learn with a recurrent neural network (RNN), we introduce a self-attention mechanism with tatum-synchronous positional encoding into the decoder. To mitigate the difficulty of training the self-attention-based model from an insufficient amount of paired data and improve the musical naturalness of the estimated scores, we propose a regularized training method that uses a global structure-aware masked language (score) model with a self-attention mechanism pretrained from an extensive collection of drum scores. Experimental results showed that the proposed regularized model outperformed the conventional RNN-based model in terms of the tatum-level error rate and the frame-level F-measure, even when only a limited amount of paired data was available so that the non-regularized model underperformed the RNN-based model.

Ryoto Ishizuka, Ryo Nishikimi, Eita Nakamura et al.

This paper describes a neural drum transcription method that detects from music signals the onset times of drums at the $\textit{tatum}$ level, where tatum times are assumed to be estimated in advance. In conventional studies on drum transcription, deep neural networks (DNNs) have often been used to take a music spectrogram as input and estimate the onset times of drums at the $\textit{frame}$ level. The major problem with such frame-to-frame DNNs, however, is that the estimated onset times do not often conform with the typical tatum-level patterns appearing in symbolic drum scores because the long-term musically meaningful structures of those patterns are difficult to learn at the frame level. To solve this problem, we propose a regularized training method for a frame-to-tatum DNN. In the proposed method, a tatum-level probabilistic language model (gated recurrent unit (GRU) network or repetition-aware bi-gram model) is trained from an extensive collection of drum scores. Given that the musical naturalness of tatum-level onset times can be evaluated by the language model, the frame-to-tatum DNN is trained with a regularizer based on the pretrained language model. The experimental results demonstrate the effectiveness of the proposed regularized training method.

Andrew McLeod, James Owers, Kazuyoshi Yoshii

In this paper, we introduce the MIDI Degradation Toolkit (MDTK), containing functions which take as input a musical excerpt (a set of notes with pitch, onset time, and duration), and return a "degraded" version of that excerpt with some error (or errors) introduced. Using the toolkit, we create the Altered and Corrupted MIDI Excerpts dataset version 1.0 (ACME v1.0), and propose four tasks of increasing difficulty to detect, classify, locate, and correct the degradations. We hypothesize that models trained for these tasks can be useful in (for example) improving automatic music transcription performance if applied as a post-processing step. To that end, MDTK includes a script that measures the distribution of different types of errors in a transcription, and creates a degraded dataset with similar properties. MDTK's degradations can also be applied dynamically to a dataset during training (with or without the above script), generating novel degraded excerpts each epoch. MDTK could also be used to test the robustness of any system designed to take MIDI (or similar) data as input (e.g. systems designed for voice separation, metrical alignment, or chord detection) to such transcription errors or otherwise noisy data. The toolkit and dataset are both publicly available online, and we encourage contribution and feedback from the community.

Kentaro Shibata, Eita Nakamura, Kazuyoshi Yoshii

We present an automatic piano transcription system that converts polyphonic audio recordings into musical scores. This has been a long-standing problem of music information processing, and recent studies have made remarkable progress in the two main component techniques: multipitch detection and rhythm quantization. Given this situation, we study a method integrating deep-neural-network-based multipitch detection and statistical-model-based rhythm quantization. In the first part, we conducted systematic evaluations and found that while the present method achieved high transcription accuracies at the note level, some global characteristics of music, such as tempo scale, metre (time signature), and bar line positions, were often incorrectly estimated. In the second part, we formulated non-local statistics of pitch and rhythmic contents that are derived from musical knowledge and studied their effects in inferring those global characteristics. We found that these statistics are markedly effective for improving the transcription results and that their optimal combination includes statistics obtained from separated hand parts. The integrated method had an overall transcription error rate of 7.1% and a downbeat F-measure of 85.6% on a dataset of popular piano music, and the generated transcriptions can be partially used for music performance and assisting human transcribers, thus demonstrating the potential for practical applications.

Yiming Wu, Tristan Carsault, Eita Nakamura et al.

This paper describes a statistically-principled semi-supervised method of automatic chord estimation (ACE) that can make effective use of music signals regardless of the availability of chord annotations. The typical approach to ACE is to train a deep classification model (neural chord estimator) in a supervised manner by using only annotated music signals. In this discriminative approach, prior knowledge about chord label sequences (model output) has scarcely been taken into account. In contrast, we propose a unified generative and discriminative approach in the framework of amortized variational inference. More specifically, we formulate a deep generative model that represents the generative process of chroma vectors (observed variables) from discrete labels and continuous features (latent variables), which are assumed to follow a Markov model favoring self-transitions and a standard Gaussian distribution, respectively. Given chroma vectors as observed data, the posterior distributions of the latent labels and features are computed approximately by using deep classification and recognition models, respectively. These three models form a variational autoencoder and can be trained jointly in a semi-supervised manner. The experimental results show that the regularization of the classification model based on the Markov prior of chord labels and the generative model of chroma vectors improved the performance of ACE even under the supervised condition. The semi-supervised learning using additional non-annotated data can further improve the performance.

Takayuki Nakatsuka, Kazuyoshi Yoshii, Yuki Koyama et al.

This paper proposes a statistical approach to 2D pose estimation from human images. The main problems with the standard supervised approach, which is based on a deep recognition (image-to-pose) model, are that it often yields anatomically implausible poses, and its performance is limited by the amount of paired data. To solve these problems, we propose a semi-supervised method that can make effective use of images with and without pose annotations. Specifically, we formulate a hierarchical generative model of poses and images by integrating a deep generative model of poses from pose features with that of images from poses and image features. We then introduce a deep recognition model that infers poses from images. Given images as observed data, these models can be trained jointly in a hierarchical variational autoencoding (image-to-pose-to-feature-to-pose-to-image) manner. The results of experiments show that the proposed reflective architecture makes estimated poses anatomically plausible, and the performance of pose estimation improved by integrating the recognition and generative models and also by feeding non-annotated images.

Tristan Carsault, Andrew McLeod, Philippe Esling et al.

This paper studies the prediction of chord progressions for jazz music by relying on machine learning models. The motivation of our study comes from the recent success of neural networks for performing automatic music composition. Although high accuracies are obtained in single-step prediction scenarios, most models fail to generate accurate multi-step chord predictions. In this paper, we postulate that this comes from the multi-scale structure of musical information and propose new architectures based on an iterative temporal aggregation of input labels. Specifically, the input and ground truth labels are merged into increasingly large temporal bags, on which we train a family of encoder-decoder networks for each temporal scale. In a second step, we use these pre-trained encoder bottleneck features at each scale in order to train a final encoder-decoder network. Furthermore, we rely on different reductions of the initial chord alphabet into three adapted chord alphabets. We perform evaluations against several state-of-the-art models and show that our multi-scale architecture outperforms existing methods in terms of accuracy and perplexity, while requiring relatively few parameters. We analyze musical properties of the results, showing the influence of downbeat position within the analysis window on accuracy, and evaluate errors using a musically-informed distance metric.

Yoshiaki Bando, Yoko Sasaki, Kazuyoshi Yoshii

This paper presents an unsupervised method that trains neural source separation by using only multichannel mixture signals. Conventional neural separation methods require a lot of supervised data to achieve excellent performance. Although multichannel methods based on spatial information can work without such training data, they are often sensitive to parameter initialization and degraded with the sources located close to each other. The proposed method uses a cost function based on a spatial model called a complex Gaussian mixture model (cGMM). This model has the time-frequency (TF) masks and direction of arrivals (DoAs) of sources as latent variables and is used for training separation and localization networks that respectively estimate these variables. This joint training solves the frequency permutation ambiguity of the spatial model in a unified deep Bayesian framework. In addition, the pre-trained network can be used not only for conducting monaural separation but also for efficiently initializing a multichannel separation algorithm. Experimental results with simulated speech mixtures showed that our method outperformed a conventional initialization method.

Eita Nakamura, Kazuyoshi Yoshii

Most work on musical score models (a.k.a. musical language models) for music transcription has focused on describing the local sequential dependence of notes in musical scores and failed to capture their global repetitive structure, which can be a useful guide for transcribing music. Focusing on rhythm, we formulate several classes of Bayesian Markov models of musical scores that describe repetitions indirectly using the sparse transition probabilities of notes or note patterns. This enables us to construct piece-specific models for unseen scores with an unfixed repetitive structure and to derive tractable inference algorithms. Moreover, to describe approximate repetitions, we explicitly incorporate a process for modifying the repeated notes/note patterns. We apply these models as prior musical score models for rhythm transcription, where piece-specific score models are inferred from performed MIDI data by Bayesian learning, in contrast to the conventional supervised construction of score models. Evaluations using the vocal melodies of popular music showed that the Bayesian models improved the transcription accuracy for most of the tested model types, indicating the universal efficacy of the proposed approach. Moreover, we found an effective data representation for modelling rhythms that maximizes the transcription accuracy and computational efficiency.

Eita Nakamura, Yasuyuki Saito, Kazuyoshi Yoshii

Automatic estimation of piano fingering is important for understanding the computational process of music performance and applicable to performance assistance and education systems. While a natural way to formulate the quality of fingerings is to construct models of the constraints/costs of performance, it is generally difficult to find appropriate parameter values for these models. Here we study an alternative data-driven approach based on statistical modeling in which the appropriateness of a given fingering is described by probabilities. Specifically, we construct two types of hidden Markov models (HMMs) and their higher-order extensions. We also study deep neural network (DNN)-based methods for comparison. Using a newly released dataset of fingering annotations, we conduct systematic evaluations of these models as well as a representative constraint-based method. We find that the methods based on high-order HMMs outperform the other methods in terms of estimation accuracies. We also quantitatively study individual difference of fingering and propose evaluation measures that can be used with multiple ground truth data. We conclude that the HMM-based methods are currently state of the art and generate acceptable fingerings in most parts and that they have certain limitations such as ignorance of phrase boundaries and interdependence of the two hands.

Kazuki Shimada, Yoshiaki Bando, Masato Mimura et al.

This paper describes multichannel speech enhancement for improving automatic speech recognition (ASR) in noisy environments. Recently, the minimum variance distortionless response (MVDR) beamforming has widely been used because it works well if the steering vector of speech and the spatial covariance matrix (SCM) of noise are given. To estimating such spatial information, conventional studies take a supervised approach that classifies each time-frequency (TF) bin into noise or speech by training a deep neural network (DNN). The performance of ASR, however, is degraded in an unknown noisy environment. To solve this problem, we take an unsupervised approach that decomposes each TF bin into the sum of speech and noise by using multichannel nonnegative matrix factorization (MNMF). This enables us to accurately estimate the SCMs of speech and noise not from observed noisy mixtures but from separated speech and noise components. In this paper we propose online MVDR beamforming by effectively initializing and incrementally updating the parameters of MNMF. Another main contribution is to comprehensively investigate the performances of ASR obtained by various types of spatial filters, i.e., time-invariant and variant versions of MVDR beamformers and those of rank-1 and full-rank multichannel Wiener filters, in combination with MNMF. The experimental results showed that the proposed method outperformed the state-of-the-art DNN-based beamforming method in unknown environments that did not match training data.

Aditya Arie Nugraha, Kouhei Sekiguchi, Kazuyoshi Yoshii

This paper proposes an approach to the joint modeling of the short-time Fourier transform magnitude and phase spectrograms with a deep generative model. We assume that the magnitude follows a Gaussian distribution and the phase follows a von Mises distribution. To improve the consistency of the phase values in the time-frequency domain, we also apply the von Mises distribution to the phase derivatives, i.e., the group delay and the instantaneous frequency. Based on these assumptions, we explore and compare several combinations of loss functions for training our models. Built upon the variational autoencoder framework, our model consists of three convolutional neural networks acting as an encoder, a magnitude decoder, and a phase decoder. In addition to the latent variables, we propose to also condition the phase estimation on the estimated magnitude. Evaluated for a time-domain speech reconstruction task, our models could generate speech with a high perceptual quality and a high intelligibility.

Kouhei Sekiguchi, Aditya Arie Nugraha, Yoshiaki Bando et al.

This paper describes a versatile method that accelerates multichannel source separation methods based on full-rank spatial modeling. A popular approach to multichannel source separation is to integrate a spatial model with a source model for estimating the spatial covariance matrices (SCMs) and power spectral densities (PSDs) of each sound source in the time-frequency domain. One of the most successful examples of this approach is multichannel nonnegative matrix factorization (MNMF) based on a full-rank spatial model and a low-rank source model. MNMF, however, is computationally expensive and often works poorly due to the difficulty of estimating the unconstrained full-rank SCMs. Instead of restricting the SCMs to rank-1 matrices with the severe loss of the spatial modeling ability as in independent low-rank matrix analysis (ILRMA), we restrict the SCMs of each frequency bin to jointly-diagonalizable but still full-rank matrices. For such a fast version of MNMF, we propose a computationally-efficient and convergence-guaranteed algorithm that is similar in form to that of ILRMA. Similarly, we propose a fast version of a state-of-the-art speech enhancement method based on a deep speech model and a low-rank noise model. Experimental results showed that the fast versions of MNMF and the deep speech enhancement method were several times faster and performed even better than the original versions of those methods, respectively.

Eita Nakamura, Kazuyoshi Yoshii

We present a statistical-modelling method for piano reduction, i.e. converting an ensemble score into piano scores, that can control performance difficulty. While previous studies have focused on describing the condition for playable piano scores, it depends on player's skill and can change continuously with the tempo. We thus computationally quantify performance difficulty as well as musical fidelity to the original score, and formulate the problem as optimization of musical fidelity under constraints on difficulty values. First, performance difficulty measures are developed by means of probabilistic generative models for piano scores and the relation to the rate of performance errors is studied. Second, to describe musical fidelity, we construct a probabilistic model integrating a prior piano-score model and a model representing how ensemble scores are likely to be edited. An iterative optimization algorithm for piano reduction is developed based on statistical inference of the model. We confirm the effect of the iterative procedure; we find that subjective difficulty and musical fidelity monotonically increase with controlled difficulty values; and we show that incorporating sequential dependence of pitches and fingering motion in the piano-score model improves the quality of reduction scores in high-difficulty cases.

Yoshiaki Bando, Masato Mimura, Katsutoshi Itoyama et al.

This paper presents a statistical method of single-channel speech enhancement that uses a variational autoencoder (VAE) as a prior distribution on clean speech. A standard approach to speech enhancement is to train a deep neural network (DNN) to take noisy speech as input and output clean speech. Although this supervised approach requires a very large amount of pair data for training, it is not robust against unknown environments. Another approach is to use non-negative matrix factorization (NMF) based on basis spectra trained on clean speech in advance and those adapted to noise on the fly. This semi-supervised approach, however, causes considerable signal distortion in enhanced speech due to the unrealistic assumption that speech spectrograms are linear combinations of the basis spectra. Replacing the poor linear generative model of clean speech in NMF with a VAE---a powerful nonlinear deep generative model---trained on clean speech, we formulate a unified probabilistic generative model of noisy speech. Given noisy speech as observed data, we can sample clean speech from its posterior distribution. The proposed method outperformed the conventional DNN-based method in unseen noisy environments.

Hiroaki Tsushima, Eita Nakamura, Katsutoshi Itoyama et al.

Generative statistical models of chord sequences play crucial roles in music processing. To capture syntactic similarities among certain chords (e.g. in C major key, between G and G7 and between F and Dm), we study hidden Markov models and probabilistic context-free grammar models with latent variables describing syntactic categories of chord symbols and their unsupervised learning techniques for inducing the latent grammar from data. Surprisingly, we find that these models often outperform conventional Markov models in predictive power, and the self-emergent categories often correspond to traditional harmonic functions. This implies the need for chord categories in harmony models from the informatics perspective.

Eita Nakamura, Kazuyoshi Yoshii, Simon Dixon

This paper presents a statistical method for use in music transcription that can estimate score times of note onsets and offsets from polyphonic MIDI performance signals. Because performed note durations can deviate largely from score-indicated values, previous methods had the problem of not being able to accurately estimate offset score times (or note values) and thus could only output incomplete musical scores. Based on observations that the pitch context and onset score times are influential on the configuration of note values, we construct a context-tree model that provides prior distributions of note values using these features and combine it with a performance model in the framework of Markov random fields. Evaluation results show that our method reduces the average error rate by around 40 percent compared to existing/simple methods. We also confirmed that, in our model, the score model plays a more important role than the performance model, and it automatically captures the voice structure by unsupervised learning.

Eita Nakamura, Kazuyoshi Yoshii, Shigeki Sagayama

In a recent conference paper, we have reported a rhythm transcription method based on a merged-output hidden Markov model (HMM) that explicitly describes the multiple-voice structure of polyphonic music. This model solves a major problem of conventional methods that could not properly describe the nature of multiple voices as in polyrhythmic scores or in the phenomenon of loose synchrony between voices. In this paper we present a complete description of the proposed model and develop an inference technique, which is valid for any merged-output HMMs for which output probabilities depend on past events. We also examine the influence of the architecture and parameters of the method in terms of accuracies of rhythm transcription and voice separation and perform comparative evaluations with six other algorithms. Using MIDI recordings of classical piano pieces, we found that the proposed model outperformed other methods by more than 12 points in the accuracy for polyrhythmic performances and performed almost as good as the best one for non-polyrhythmic performances. This reveals the state-of-the-art methods of rhythm transcription for the first time in the literature. Publicly available source codes are also provided for future comparisons.

Yukara Ikemiya, Katsutoshi Itoyama, Kazuyoshi Yoshii

This paper presents a new method of singing voice analysis that performs mutually-dependent singing voice separation and vocal fundamental frequency (F0) estimation. Vocal F0 estimation is considered to become easier if singing voices can be separated from a music audio signal, and vocal F0 contours are useful for singing voice separation. This calls for an approach that improves the performance of each of these tasks by using the results of the other. The proposed method first performs robust principal component analysis (RPCA) for roughly extracting singing voices from a target music audio signal. The F0 contour of the main melody is then estimated from the separated singing voices by finding the optimal temporal path over an F0 saliency spectrogram. Finally, the singing voices are separated again more accurately by combining a conventional time-frequency mask given by RPCA with another mask that passes only the harmonic structures of the estimated F0s. Experimental results showed that the proposed method significantly improved the performances of both singing voice separation and vocal F0 estimation. The proposed method also outperformed all the other methods of singing voice separation submitted to an international music analysis competition called MIREX 2014.