Rong Gong

Dushyant Sharma, Rong Gong, James Fosburgh et al.

We present a novel multi-channel front-end based on channel shortening with theWeighted Prediction Error (WPE) method followed by a fixed MVDR beamformer used in combination with a recently proposed self-attention-based channel combination (SACC) scheme, for tackling the distant ASR problem. We show that the proposed system used as part of a ContextNet based end-to-end (E2E) ASR system outperforms leading ASR systems as demonstrated by a 21.6% reduction in relative WER on a multi-channel LibriSpeech playback dataset. We also show how dereverberation prior to beamforming is beneficial and compare the WPE method with a modified neural channel shortening approach. An analysis of the non-intrusive estimate of the signal C50 confirms that the 8 channel WPE method provides significant dereverberation of the signals (13.6 dB improvement). We also show how the weights of the SACC system allow the extraction of accurate spatial information which can be beneficial for other speech processing applications like diarization.

Rong Gong, Hongfei Xue, Lezhi Wang et al.

The rapid advancements in speech technologies over the past two decades have led to human-level performance in tasks like automatic speech recognition (ASR) for fluent speech. However, the efficacy of these models diminishes when applied to atypical speech, such as stuttering. This paper introduces AS-70, the first publicly available Mandarin stuttered speech dataset, which stands out as the largest dataset in its category. Encompassing conversational and voice command reading speech, AS-70 includes verbatim manual transcription, rendering it suitable for various speech-related tasks. Furthermore, baseline systems are established, and experimental results are presented for ASR and stuttering event detection (SED) tasks. By incorporating this dataset into the model fine-tuning, significant improvements in the state-of-the-art ASR models, e.g., Whisper and Hubert, are observed, enhancing their inclusivity in addressing stuttered speech.

Rong Gong, Carl Quillen, Dushyant Sharma et al.

When a sufficiently large far-field training data is presented, jointly optimizing a multichannel frontend and an end-to-end (E2E) Automatic Speech Recognition (ASR) backend shows promising results. Recent literature has shown traditional beamformer designs, such as MVDR (Minimum Variance Distortionless Response) or fixed beamformers can be successfully integrated as the frontend into an E2E ASR system with learnable parameters. In this work, we propose the self-attention channel combinator (SACC) ASR frontend, which leverages the self-attention mechanism to combine multichannel audio signals in the magnitude spectral domain. Experiments conducted on a multichannel playback test data shows that the SACC achieved a 9.3% WERR compared to a state-of-the-art fixed beamformer-based frontend, both jointly optimized with a ContextNet-based ASR backend. We also demonstrate the connection between the SACC and the traditional beamformers, and analyze the intermediate outputs of the SACC.

Eduardo Fonseca, Rong Gong, Xavier Serra

In the past, Acoustic Scene Classification systems have been based on hand crafting audio features that are input to a classifier. Nowadays, the common trend is to adopt data driven techniques, e.g., deep learning, where audio representations are learned from data. In this paper, we propose a system that consists of a simple fusion of two methods of the aforementioned types: a deep learning approach where log-scaled mel-spectrograms are input to a convolutional neural network, and a feature engineering approach, where a collection of hand-crafted features is input to a gradient boosting machine. We first show that both methods provide complementary information to some extent. Then, we use a simple late fusion strategy to combine both methods. We report classification accuracy of each method individually and the combined system on the TUT Acoustic Scenes 2017 dataset. The proposed fused system outperforms each of the individual methods and attains a classification accuracy of 72.8% on the evaluation set, improving the baseline system by 11.8%.

Rong Gong, Xavier Serra

In this paper, we propose an efficient and reproducible deep learning model for musical onset detection (MOD). We first review the state-of-the-art deep learning models for MOD, and identify their shortcomings and challenges: (i) the lack of hyper-parameter tuning details, (ii) the non-availability of code for training models on other datasets, and (iii) ignoring the network capability when comparing different architectures. Taking the above issues into account, we experiment with seven deep learning architectures. The most efficient one achieves equivalent performance to our implementation of the state-of-the-art architecture. However, it has only 28.3% of the total number of trainable parameters compared to the state-of-the-art. Our experiments are conducted using two different datasets: one mainly consists of instrumental music excerpts, and another developed by ourselves includes only solo singing voice excerpts. Further, inter-dataset transfer learning experiments are conducted. The results show that the model pre-trained on one dataset fails to detect onsets on another dataset, which denotes the importance of providing the implementation code to enable re-training the model for a different dataset. Datasets, code and a Jupyter notebook running on Google Colab are publicly available to make this research understandable and easy to reproduce.

Rong Gong, Xavier Serra

In this paper, we tackle the singing voice phoneme segmentation problem in the singing training scenario by using language-independent information -- onset and prior coarse duration. We propose a two-step method. In the first step, we jointly calculate the syllable and phoneme onset detection functions (ODFs) using a convolutional neural network (CNN). In the second step, the syllable and phoneme boundaries and labels are inferred hierarchically by using a duration-informed hidden Markov model (HMM). To achieve the inference, we incorporate the a priori duration model as the transition probabilities and the ODFs as the emission probabilities into the HMM. The proposed method is designed in a language-independent way such that no phoneme class labels are used. For the model training and algorithm evaluation, we collect a new jingju (also known as Beijing or Peking opera) solo singing voice dataset and manually annotate the boundaries and labels at phrase, syllable and phoneme levels. The dataset is publicly available. The proposed method is compared with a baseline method based on hidden semi-Markov model (HSMM) forced alignment. The evaluation results show that the proposed method outperforms the baseline by a large margin regarding both segmentation and onset detection tasks.

Rong Gong, Xavier Serra

Music Information Retrieval (MIR) technologies have been proven useful in assisting western classical singing training. Jingju (also known as Beijing or Peking opera) singing is different from western singing in terms of most of the perceptual dimensions, and the trainees are taught by using mouth/heart method. In this paper, we first present the training method used in the professional jingju training classroom scenario and show the potential benefits of introducing the MIR technologies into the training process. The main part of this paper dedicates to identify the potential MIR technologies for jingju singing training. To this intent, we answer the question: how the jingju singing tutors and trainees value the importance of each jingju musical dimension-intonation, rhythm, loudness, tone quality and pronunciation? This is done by (i) classifying the classroom singing practices, tutor's verbal feedbacks into these 5 dimensions, (ii) surveying the trainees. Then, with the help of the music signal analysis, a finer inspection on the classroom practice recording examples reveals the detailed elements in the training process. Finally, based on the above analysis, several potential MIR technologies are identified and would be useful for the jingju singing training.

Rong Gong, Rafael Caro Repetto, Xavier Serra

The data-driven computational research on automatic jingju (also known as Beijing or Peking opera) singing evaluation lacks a suitable and comprehensive a cappella singing audio dataset. In this work, we present an a cappella singing audio dataset which consists of 120 arias, accounting for 1265 melodic lines. This dataset is also an extension our existing CompMusic jingju corpus. Both professional and amateur singers were invited to the dataset recording sessions, and the most common jingju musical elements have been covered. This dataset is also accompanied by metadata per aria and melodic line annotated for automatic singing evaluation research purpose. All the gathered data is openly available online.

Rong Gong, Jordi Pons, Xavier Serra

We approach the singing phrase audio to score matching problem by using phonetic and duration information - with a focus on studying the jingju a cappella singing case. We argue that, due to the existence of a basic melodic contour for each mode in jingju music, only using melodic information (such as pitch contour) will result in an ambiguous matching. This leads us to propose a matching approach based on the use of phonetic and duration information. Phonetic information is extracted with an acoustic model shaped with our data, and duration information is considered with the Hidden Markov Models (HMMs) variants we investigate. We build a model for each lyric path in our scores and we achieve the matching by ranking the posterior probabilities of the decoded most likely state sequences. Three acoustic models are investigated: (i) convolutional neural networks (CNNs), (ii) deep neural networks (DNNs) and (iii) Gaussian mixture models (GMMs). Also, two duration models are compared: (i) hidden semi-Markov model (HSMM) and (ii) post-processor duration model. Results show that CNNs perform better in our (small) audio dataset and also that HSMM outperforms the post-processor duration model.

Jordi Pons, Rong Gong, Xavier Serra

This paper introduces a new score-informed method for the segmentation of jingju a cappella singing phrase into syllables. The proposed method estimates the most likely sequence of syllable boundaries given the estimated syllable onset detection function (ODF) and its score. Throughout the paper, we first examine the jingju syllables structure and propose a definition of the term "syllable onset". Then, we identify which are the challenges that jingju a cappella singing poses. Further, we investigate how to improve the syllable ODF estimation with convolutional neural networks (CNNs). We propose a novel CNN architecture that allows to efficiently capture different time-frequency scales for estimating syllable onsets. In addition, we propose using a score-informed Viterbi algorithm -instead of thresholding the onset function-, because the available musical knowledge we have (the score) can be used to inform the Viterbi algorithm in order to overcome the identified challenges. The proposed method outperforms the state-of-the-art in syllable segmentation for jingju a cappella singing. We further provide an analysis of the segmentation errors which points possible research directions.

Jordi Pons, Olga Slizovskaia, Rong Gong et al.

The focus of this work is to study how to efficiently tailor Convolutional Neural Networks (CNNs) towards learning timbre representations from log-mel magnitude spectrograms. We first review the trends when designing CNN architectures. Through this literature overview we discuss which are the crucial points to consider for efficiently learning timbre representations using CNNs. From this discussion we propose a design strategy meant to capture the relevant time-frequency contexts for learning timbre, which permits using domain knowledge for designing architectures. In addition, one of our main goals is to design efficient CNN architectures -- what reduces the risk of these models to over-fit, since CNNs' number of parameters is minimized. Several architectures based on the design principles we propose are successfully assessed for different research tasks related to timbre: singing voice phoneme classification, musical instrument recognition and music auto-tagging.