Search

Neural audio codec tokens serve as the fundamental building blocks for speech language model (SLM)-based speech generation. However, there is no systematic understanding on how the codec system affects the speech generation performance of the SLM. In this work, we examine codec tokens within SLM framework for speech generation to provide insights for effective codec design. We retrain existing high-performing neural codec models on the same data set and loss functions to compare their performance in a uniform setting. We integrate codec tokens into two SLM systems: masked-based parallel speech generation system and an auto-regressive (AR) plus non-auto-regressive (NAR) model-based system. Our findings indicate that better speech reconstruction in codec systems does not guarantee improved speech generation in SLM. A high-quality codec decoder is crucial for natural speech production in SLM, while speech intelligibility depends more on quantization mechanism., Comment: Accepted by SLT-2024

The Listening in Spatialized Noise Sentences (LiSN-S) is a test to evaluate auditory spatial processing currently only available in the English language. It produces a three-dimensional auditory environment under headphones and uses a simple repetition response protocol to determine speech reception thresholds (SRTs) for sentences presented in competing speech under various conditions. In order to develop the LiSN-S test in Brazilian Portuguese, it was necessary to prepare a speech database recorded by professional voice actresses and to devise presentation software. These sentences were presented to 35 adults (aged between 19 and 40 years) and 24 children (aged between 8 and 10 years), all with normal hearing-verified through tone and speech audiometry and tympanometry-and good performance at school. We used a logistic curve describing word error rate versus presentation level, fitted for each sentence, to select a set of 120 sentences for the test. Furthermore, all selected sentences were adjusted in amplitude for equal intelligibility. The framework of LiSN-S in Brazilian Portuguese is ready for normative data analysis. After its conclusion, we believe it will contribute to diagnosing and rehabilitating Brazilian children with complaints related to hearing difficulties in noisy environments

Speech emotion recognition (SER) classifies human emotions in speech with a computer model. Recently, performance in SER has steadily increased as deep learning techniques have adapted. However, unlike many domains that use speech data, data for training in the SER model is insufficient. This causes overfitting of training of the neural network, resulting in performance degradation. In fact, successful emotion recognition requires an effective preprocessing method and a model structure that efficiently uses the number of weight parameters. In this study, we propose using eight dataset versions with different frequency-time resolutions to search for an effective emotional speech preprocessing method. We propose a 6-layer convolutional neural network (CNN) model with efficient channel attention (ECA) to pursue an efficient model structure. In particular, the well-positioned ECA blocks can improve channel feature representation with only a few parameters. With the interactive emotional dyadic motion capture (IEMOCAP) dataset, increasing the frequency resolution in preprocessing emotional speech can improve emotion recognition performance. Also, ECA after the deep convolution layer can effectively increase channel feature representation. Consequently, the best result (79.37UA 79.68WA) can be obtained, exceeding the performance of previous SER models. Furthermore, to compensate for the lack of emotional speech data, we experiment with multiple preprocessing data methods that augment trainable data preprocessed with all different settings from one sample. In the experiment, we can achieve the highest result (80.28UA 80.46WA).

This paper introduces MetaBGM, a groundbreaking framework for generating background music that adapts to dynamic scenes and real-time user interactions. We define multi-scene as variations in environmental contexts, such as transitions in game settings or movie scenes. To tackle the challenge of converting backend data into music description texts for audio generation models, MetaBGM employs a novel two-stage generation approach that transforms continuous scene and user state data into these texts, which are then fed into an audio generation model for real-time soundtrack creation. Experimental results demonstrate that MetaBGM effectively generates contextually relevant and dynamic background music for interactive applications.

Speech tokenization serves as the foundation of speech language model (LM), enabling them to perform various tasks such as spoken language modeling, text-to-speech, speech-to-text, etc. Most speech tokenizers are trained independently of the LM training process, relying on separate acoustic models and quantization methods. Following such an approach may create a mismatch between the tokenization process and its usage afterward. In this study, we propose a novel approach to training a speech tokenizer by leveraging objectives from pre-trained textual LMs. We advocate for the integration of this objective into the process of learning discrete speech representations. Our aim is to transform features from a pre-trained speech model into a new feature space that enables better clustering for speech LMs. We empirically investigate the impact of various model design choices, including speech vocabulary size and text LM size. Our results demonstrate the proposed tokenization method outperforms the evaluated baselines considering both spoken language modeling and speech-to-text. More importantly, unlike prior work, the proposed method allows the utilization of a single pre-trained LM for processing both speech and text inputs, setting it apart from conventional tokenization approaches.

This paper presents the Multimodal Analyzing System for Laryngoscope (MASL), a system that combines audio and video data to automatically extract key segments and metrics from laryngeal videostroboscopic videos for clinical assessment. MASL integrates glottis detection with keyword spotting to analyze patient vocalizations and refine video highlights for better inspection of vocal cord movements. The system includes a strobing video extraction module that identifies frames by analyzing hue, saturation, and value fluctuations. MASL also provides effective metrics for vocal cord paralysis detection, employing a two-stage glottis segmentation process using U-Net followed by diffusion-based refinement to reduce false positives. Instead of glottal area waveforms, MASL estimates anterior glottic angle waveforms (AGAW) from glottis masks, evaluating both left and right vocal cords to detect unilateral vocal cord paralysis (UVFP). By comparing AGAW variances, MASL distinguishes between left and right paralysis. Ablation studies and experiments on public and real-world datasets validate MASL's segmentation module and demonstrate its ability to provide reliable metrics for UVFP diagnosis.

We present aTENNuate, a simple deep state-space autoencoder configured for efficient online raw speech enhancement in an end-to-end fashion. The network's performance is primarily evaluated on raw speech denoising, with additional assessments on tasks such as super-resolution and de-quantization. We benchmark aTENNuate on the VoiceBank + DEMAND and the Microsoft DNS1 synthetic test sets. The network outperforms previous real-time denoising models in terms of PESQ score, parameter count, MACs, and latency. Even as a raw waveform processing model, the model maintains high fidelity to the clean signal with minimal audible artifacts. In addition, the model remains performant even when the noisy input is compressed down to 4000Hz and 4 bits, suggesting general speech enhancement capabilities in low-resource environments., Comment: 7 pages, 2 figures

Measuring 3D geometric structures of indoor scenes requires dedicated depth sensors, which are not always available. Echo-based depth estimation has recently been studied as a promising alternative solution. All previous studies have assumed the use of echoes in the audible range. However, one major problem is that audible echoes cannot be used in quiet spaces or other situations where producing audible sounds is prohibited. In this paper, we consider echo-based depth estimation using inaudible ultrasonic echoes. While ultrasonic waves provide high measurement accuracy in theory, the actual depth estimation accuracy when ultrasonic echoes are used has remained unclear, due to its disadvantage of being sensitive to noise and susceptible to attenuation. We first investigate the depth estimation accuracy when the frequency of the sound source is restricted to the high-frequency band, and found that the accuracy decreased when the frequency was limited to ultrasonic ranges. Based on this observation, we propose a novel deep learning method to improve the accuracy of ultrasonic echo-based depth estimation by using audible echoes as auxiliary data only during training. Experimental results with a public dataset demonstrate that our method improves the estimation accuracy., Comment: ICIP 2024

This work proposes FireRedTTS, a foundation text-to-speech framework, to meet the growing demands for personalized and diverse generative speech applications. The framework comprises three parts: data processing, foundation system, and downstream applications. First, we comprehensively present our data processing pipeline, which transforms massive raw audio into a large-scale high-quality TTS dataset with rich annotations and a wide coverage of content, speaking style, and timbre. Then, we propose a language-model-based foundation TTS system. The speech signal is compressed into discrete semantic tokens via a semantic-aware speech tokenizer, and can be generated by a language model from the prompt text and audio. Then, a two-stage waveform generator is proposed to decode them to the high-fidelity waveform. We present two applications of this system: voice cloning for dubbing and human-like speech generation for chatbots. The experimental results demonstrate the solid in-context learning capability of FireRedTTS, which can stably synthesize high-quality speech consistent with the prompt text and audio. For dubbing, FireRedTTS can clone target voices in a zero-shot way for the UGC scenario and adapt to studio-level expressive voice characters in the PUGC scenario via few-shot fine-tuning with 1-hour recording. Moreover, FireRedTTS achieves controllable human-like speech generation in a casual style with paralinguistic behaviors and emotions via instruction tuning, to better serve spoken chatbots.

Progress in the task of symbolic music generation may be lagging behind other tasks like audio and text generation, in part because of the scarcity of symbolic training data. In this paper, we leverage the greater scale of audio music data by applying pre-trained MIR models (for transcription, beat tracking, structure analysis, etc.) to extract symbolic events and encode them into token sequences. To the best of our knowledge, this work is the first to demonstrate the feasibility of training symbolic generation models solely from auto-transcribed audio data. Furthermore, to enhance the controllability of the trained model, we introduce SymPAC (Symbolic Music Language Model with Prompting And Constrained Generation), which is distinguished by using (a) prompt bars in encoding and (b) a technique called Constrained Generation via Finite State Machines (FSMs) during inference time. We show the flexibility and controllability of this approach, which may be critical in making music AI useful to creators and users., Comment: ISMIR 2024

The advancements in audio generative models have opened up new challenges in their responsible disclosure and the detection of their misuse. In response, we introduce a method to watermark latent generative models by a specific watermarking of their training data. The resulting watermarked models produce latent representations whose decoded outputs are detected with high confidence, regardless of the decoding method used. This approach enables the detection of the generated content without the need for a post-hoc watermarking step. It provides a more secure solution for open-sourced models and facilitates the identification of derivative works that fine-tune or use these models without adhering to their license terms. Our results indicate for instance that generated outputs are detected with an accuracy of more than 75% at a false positive rate of $10^{-3}$, even after fine-tuning the latent generative model.

Diffusion models have shown promising results in cross-modal generation tasks involving audio and music, such as text-to-sound and text-to-music generation. These text-controlled music generation models typically focus on generating music by capturing global musical attributes like genre and mood. However, music composition is a complex, multilayered task that often involves musical arrangement as an integral part of the process. This process involves composing each instrument to align with existing ones in terms of beat, dynamics, harmony, and melody, requiring greater precision and control over tracks than text prompts usually provide. In this work, we address these challenges by extending the MusicLDM, a latent diffusion model for music, into a multi-track generative model. By learning the joint probability of tracks sharing a context, our model is capable of generating music across several tracks that correspond well to each other, either conditionally or unconditionally. Additionally, our model is capable of arrangement generation, where the model can generate any subset of tracks given the others (e.g., generating a piano track complementing given bass and drum tracks). We compared our model with an existing multi-track generative model and demonstrated that our model achieves considerable improvements across objective metrics for both total and arrangement generation tasks.

Objective: The Daily Phonotrauma Index (DPI) can quantify pathophysiological mechanisms associated with daily voice use in individuals with phonotraumatic vocal hyperfunction (PVH). Since DPI was developed based on week-long ambulatory voice monitoring, this study investigated if DPI can achieve comparable performance using (1) short laboratory speech tasks and (2) fewer than seven days of ambulatory data. Method: An ambulatory voice monitoring system recorded the vocal function/behavior of 134 females with PVH and vocally healthy matched controls in two different conditions. In the lab, the participants read the first paragraph of the Rainbow Passage and produced spontaneous speech (in-lab data). They were then monitored for seven days (in-field data). Separate DPI models were trained from in-lab and in-field data using the standard deviation of the difference between the magnitude of the first two harmonics (H1-H2) and the skewness of neck-surface acceleration magnitude. First, 10-fold cross-validation evaluated classification performance of in-lab and in-field DPIs. Second, the effect of the number of ambulatory monitoring days on the accuracy of in-field DPI classification was quantified. Results: The average in-lab DPI accuracy computed from the Rainbow passage and spontaneous speech were, respectively, 57.9% and 48.9%, which are close to chance performance. The average classification accuracy of in-field DPI was significantly higher with a very large effect size (73.4%, Cohens D = 1.8). Second, the average in-field DPI accuracy increased from 66.5% for one day to 75.0% for seven days, with the gain of including an additional day on accuracy dropping below 1 percentage point after 4 days., Comment: The paper is submitted to JSLHR

Target speaker extraction aims to isolate the voice of a specific speaker from mixed speech. Traditionally, this process has relied on extracting a speaker embedding from a reference speech, necessitating a speaker recognition model. However, identifying an appropriate speaker recognition model can be challenging, and using the target speaker embedding as reference information may not be optimal for target speaker extraction tasks. This paper introduces a Universal Speaker Embedding-Free Target Speaker Extraction (USEF-TSE) framework that operates without relying on speaker embeddings. USEF-TSE utilizes a multi-head cross-attention mechanism as a frame-level target speaker feature extractor. This innovative approach allows mainstream speaker extraction solutions to bypass the dependency on speaker recognition models and to fully leverage the information available in the enrollment speech, including speaker characteristics and contextual details. Additionally, USEF-TSE can seamlessly integrate with any time-domain or time-frequency domain speech separation model to achieve effective speaker extraction. Experimental results show that our proposed method achieves state-of-the-art (SOTA) performance in terms of Scale-Invariant Signal-to-Distortion Ratio (SI-SDR) on the WSJ0-2mix, WHAM!, and WHAMR! datasets, which are standard benchmarks for monaural anechoic, noisy and noisy-reverberant two-speaker speech separation and speaker extraction., Comment: 13 pages, 6 figures

Self-Supervised Learning (SSL) has proven to be effective in various domains, including speech processing. However, SSL is computationally and memory expensive. This is in part due the quadratic complexity of multi-head self-attention (MHSA). Alternatives for MHSA have been proposed and used in the speech domain, but have yet to be investigated properly in an SSL setting. In this work, we study the effects of replacing MHSA with recent state-of-the-art alternatives that have linear complexity, namely, HyperMixing, Fastformer, SummaryMixing, and Mamba. We evaluate these methods by looking at the speed, the amount of VRAM consumed, and the performance on the SSL MP3S benchmark. Results show that these linear alternatives maintain competitive performance compared to MHSA while, on average, decreasing VRAM consumption by around 20% to 60% and increasing speed from 7% to 65% for input sequences ranging from 20 to 80 seconds., Comment: Accepted in the IEEE Soken Language Technology Workshop 2024

Continuous speech can be converted into a discrete sequence by deriving discrete units from the hidden features of self-supervised learned (SSL) speech models. Although SSL models are becoming larger and trained on more data, they are often sensitive to real-life distortions like additive noise or reverberation, which translates to a shift in discrete units. We propose a parameter-efficient approach to generate noise-robust discrete units from pre-trained SSL models by training a small encoder-decoder model, with or without adapters, to simultaneously denoise and discretise the hidden features of the SSL model. The model learns to generate a clean discrete sequence for a noisy utterance, conditioned on the SSL features. The proposed denoiser outperforms several pre-training methods on the tasks of noisy discretisation and noisy speech recognition, and can be finetuned to the target environment with a few recordings of unlabeled target data., Comment: Accepted at SLT2024

While universal vocoders have achieved proficient waveform generation across diverse voices, their integration into text-to-speech (TTS) tasks often results in degraded synthetic quality. To address this challenge, we present a novel augmentation technique for training universal vocoders. Our training scheme randomly applies linear smoothing filters to input acoustic features, facilitating vocoder generalization across a wide range of smoothings. It significantly mitigates the training-inference mismatch, enhancing the naturalness of synthetic output even when the acoustic model produces overly smoothed features. Notably, our method is applicable to any vocoder without requiring architectural modifications or dependencies on specific acoustic models. The experimental results validate the superiority of our vocoder over conventional methods, achieving 11.99% and 12.05% improvements in mean opinion scores when integrated with Tacotron 2 and FastSpeech 2 TTS acoustic models, respectively., Comment: 4 pages, 4 figures, for demo samples, see https://sytronik.github.io/demos/voc_smth_aug/

In the study of auditory attention, it has been revealed that there exists a robust correlation between attended speech and elicited neural responses, measurable through electroencephalography (EEG). Therefore, it is possible to use the attention information available within EEG signals to guide the extraction of the target speaker in a cocktail party computationally. In this paper, we present a neuro-guided speaker extraction model, i.e. NeuroSpex, using the EEG response of the listener as the sole auxiliary reference cue to extract attended speech from monaural speech mixtures. We propose a novel EEG signal encoder that captures the attention information. Additionally, we propose a cross-attention (CA) mechanism to enhance the speech feature representations, generating a speaker extraction mask. Experimental results on a publicly available dataset demonstrate that our proposed model outperforms two baseline models across various evaluation metrics.

Psychotherapy or counseling is typically conducted through spoken conversation between a therapist and a client. Analyzing the speech characteristics of psychotherapeutic interactions can help understand the factors associated with effective psychotherapy. This paper introduces CUEMPATHY, a large-scale speech dataset collected from actual counseling sessions. The dataset consists of 156 counseling sessions involving 39 therapist-client dyads. The process of speech data collection, subjective ratings (one observer and two client ratings), and transcription are described. An automatic speech and text processing system is developed to locate the time stamps of speaker turns in each session. Examining the relationships among the three subjective ratings suggests that observer and client ratings have no significant correlation, while the client-rated measures are significantly correlated. The intensity similarity between the therapist and the client, measured by the averaged absolute difference of speaker-turn-level intensities, is associated with the psychotherapy outcomes. Recent studies on the acoustic and linguistic characteristics of the CUEMPATHY are introduced., Comment: Accepted by ISCSLP 2022

Articulatory trajectories like electromagnetic articulography (EMA) provide a low-dimensional representation of the vocal tract filter and have been used as natural, grounded features for speech synthesis. Differentiable digital signal processing (DDSP) is a parameter-efficient framework for audio synthesis. Therefore, integrating low-dimensional EMA features with DDSP can significantly enhance the computational efficiency of speech synthesis. In this paper, we propose a fast, high-quality, and parameter-efficient DDSP articulatory vocoder that can synthesize speech from EMA, F0, and loudness. We incorporate several techniques to solve the harmonics / noise imbalance problem, and add a multi-resolution adversarial loss for better synthesis quality. Our model achieves a transcription word error rate (WER) of 6.67% and a mean opinion score (MOS) of 3.74, with an improvement of 1.63% and 0.16 compared to the state-of-the-art (SOTA) baseline. Our DDSP vocoder is 4.9x faster than the baseline on CPU during inference, and can generate speech of comparable quality with only 0.4M parameters, in contrast to the 9M parameters required by the SOTA., Comment: accepted for Spoken Language Technology Workshop 2024

In recent years, deep learning has significantly advanced the MIDI domain, solidifying music generation as a key application of artificial intelligence. However, existing research primarily focuses on Western music and encounters challenges in generating melodies for Chinese traditional music, especially in capturing modal characteristics and emotional expression. To address these issues, we propose a new architecture, the Dual-Feature Modeling Module, which integrates the long-range dependency modeling of the Mamba Block with the global structure capturing capabilities of the Transformer Block. Additionally, we introduce the Bidirectional Mamba Fusion Layer, which integrates local details and global structures through bidirectional scanning, enhancing the modeling of complex sequences. Building on this architecture, we propose the REMI-M representation, which more accurately captures and generates modal information in melodies. To support this research, we developed FolkDB, a high-quality Chinese traditional music dataset encompassing various styles and totaling over 11 hours of music. Experimental results demonstrate that the proposed architecture excels in generating melodies with Chinese traditional music characteristics, offering a new and effective solution for music generation.

Due to the inherent difficulty in modeling phonetic similarities across different languages, code-switching speech recognition presents a formidable challenge. This study proposes a Collaborative-MoE, a Mixture of Experts (MoE) model that leverages a collaborative mechanism among expert groups. Initially, a preceding routing network explicitly learns Language Identification (LID) tasks and selects experts based on acquired LID weights. This process ensures robust routing information to the MoE layer, mitigating interference from diverse language domains on expert network parameter updates. The LID weights are also employed to facilitate inter-group collaboration, enabling the integration of language-specific representations. Furthermore, within each language expert group, a gating network operates unsupervised to foster collaboration on attributes beyond language. Extensive experiments demonstrate the efficacy of our approach, achieving significant performance enhancements compared to alternative methods. Importantly, our method preserves the efficient inference capabilities characteristic of MoE models without necessitating additional pre-training., Comment: Accepted by IEEE SLT 2024

This technical report outlines our submission system for the CHiME-8 NOTSOFAR-1 Challenge. The primary difficulty of this challenge is the dataset recorded across various conference rooms, which captures real-world complexities such as high overlap rates, background noises, a variable number of speakers, and natural conversation styles. To address these issues, we optimized the system in several aspects: For front-end speech signal processing, we introduced a data-driven joint training method for diarization and separation (JDS) to enhance audio quality. Additionally, we also integrated traditional guided source separation (GSS) for multi-channel track to provide complementary information for the JDS. For back-end speech recognition, we enhanced Whisper with WavLM, ConvNeXt, and Transformer innovations, applying multi-task training and Noise KLD augmentation, to significantly advance ASR robustness and accuracy. Our system attained a Time-Constrained minimum Permutation Word Error Rate (tcpWER) of 14.265% and 22.989% on the CHiME-8 NOTSOFAR-1 Dev-set-2 multi-channel and single-channel tracks, respectively.

We propose a new speech discrete token vocoder, vec2wav 2.0, which advances voice conversion (VC). We use discrete tokens from speech self-supervised models as the content features of source speech, and treat VC as a prompted vocoding task. To amend the loss of speaker timbre in the content tokens, vec2wav 2.0 utilizes the WavLM features to provide strong timbre-dependent information. A novel adaptive Snake activation function is proposed to better incorporate timbre into the waveform reconstruction process. In this way, vec2wav 2.0 learns to alter the speaker timbre appropriately given different reference prompts. Also, no supervised data is required for vec2wav 2.0 to be effectively trained. Experimental results demonstrate that vec2wav 2.0 outperforms all other baselines to a considerable margin in terms of audio quality and speaker similarity in any-to-any VC. Ablation studies verify the effects made by the proposed techniques. Moreover, vec2wav 2.0 achieves competitive cross-lingual VC even only trained on monolingual corpus. Thus, vec2wav 2.0 shows timbre can potentially be manipulated only by speech token vocoders, pushing the frontiers of VC and speech synthesis., Comment: 5 pages, 4 figures

In recent years, artificial intelligence (AI) has made significant progress in the field of music generation, driving innovation in music creation and applications. This paper provides a systematic review of the latest research advancements in AI music generation, covering key technologies, models, datasets, evaluation methods, and their practical applications across various fields. The main contributions of this review include: (1) presenting a comprehensive summary framework that systematically categorizes and compares different technological approaches, including symbolic generation, audio generation, and hybrid models, helping readers better understand the full spectrum of technologies in the field; (2) offering an extensive survey of current literature, covering emerging topics such as multimodal datasets and emotion expression evaluation, providing a broad reference for related research; (3) conducting a detailed analysis of the practical impact of AI music generation in various application domains, particularly in real-time interaction and interdisciplinary applications, offering new perspectives and insights; (4) summarizing the existing challenges and limitations of music quality evaluation methods and proposing potential future research directions, aiming to promote the standardization and broader adoption of evaluation techniques. Through these innovative summaries and analyses, this paper serves as a comprehensive reference tool for researchers and practitioners in AI music generation, while also outlining future directions for the field.

Indonesian and Western gamelan ensembles are investigated with respect to performance differences. Thereby, the often exotistic history of this music in the West might be reflected in contemporary tonal system, articulation, or large-scale form differences. Analyzing recordings of four Western and five Indonesian orchestras with respect to tonal systems and timbre features and using self-organizing Kohonen map (SOM) as a machine learning algorithm, a clear clustering between Indonesian and Western ensembles appears using certain psychoacoustic features. These point to a reduced articulation and large-scale form variability of Western ensembles compared to Indonesian ones. The SOM also clusters the ensembles with respect to their tonal systems, but no clusters between Indonesian and Western ensembles can be found in this respect. Therefore, a clear analogy between lower articulatory variability and large-scale form variation and a more exostistic, mediative and calm performance expectation and reception of gamelan in the West therefore appears., Comment: 5 figures, 4 tables

We address a practical scenario of anomaly detection for industrial sound data, where the sound of a target machine is corrupted by background noise and interference from neighboring machines. Overcoming this challenge is difficult since the interference is often virtually indistinguishable from the target machine without additional information. To address the issue, we propose SSAD, a framework of source separation (SS) followed by anomaly detection (AD), which leverages machine activity information, often readily available in practical settings. SSAD consists of two components: (i) activity-informed SS, enabling effective source separation even given interference with similar timbre, and (ii) two-step masking, robustifying anomaly detection by emphasizing anomalies aligned with the machine activity. Our experiments demonstrate that SSAD achieves comparable accuracy to a baseline with full access to clean signals, while SSAD is provided only a corrupted signal and activity information. In addition, thanks to the activity-informed SS and AD with the two-step masking, SSAD outperforms standard approaches, particularly in cases with interference. It highlights the practical efficacy of SSAD in addressing the complexities of anomaly detection in industrial sound data., Comment: Thsis is an extended version of https://ieeexplore.ieee.org/document/10095113

In this work, we explore the use and reliability of Large Language Models (LLMs) in musicology. From a discussion with experts and students, we assess the current acceptance and concerns regarding this, nowadays ubiquitous, technology. We aim to go one step further, proposing a semi-automatic method to create an initial benchmark using retrieval-augmented generation models and multiple-choice question generation, validated by human experts. Our evaluation on 400 human-validated questions shows that current vanilla LLMs are less reliable than retrieval augmented generation from music dictionaries. This paper suggests that the potential of LLMs in musicology requires musicology driven research that can specialized LLMs by including accurate and reliable domain knowledge.

In this study, we investigate if noise-augmented training can concurrently improve adversarial robustness in automatic speech recognition (ASR) systems. We conduct a comparative analysis of the adversarial robustness of four different state-of-the-art ASR architectures, where each of the ASR architectures is trained under three different augmentation conditions: one subject to background noise, speed variations, and reverberations, another subject to speed variations only, and a third without any form of data augmentation. The results demonstrate that noise augmentation not only improves model performance on noisy speech but also the model's robustness to adversarial attacks.

Accurately estimating the direction-of-arrival (DOA) of a speech source using a compact microphone array (CMA) is often complicated by background noise and reverberation. A commonly used DOA estimation method is the steered response power with phase transform (SRP-PHAT) function, which has been shown to work reliably in moderate levels of noise and reverberation. Since for closely spaced microphones the spatial coherence of noise and reverberation may be high over an extended frequency range, this may negatively affect the SRP-PHAT spectra, resulting in DOA estimation errors. Assuming the availability of an auxiliary microphone at an unknown position which is spatially separated from the CMA, in this paper we propose to compute the SRP-PHAT spectra between the microphones of the CMA based on the SRP-PHAT spectra between the auxiliary microphone and the microphones of the CMA. For different levels of noise and reverberation, we show how far the auxiliary microphone needs to be spatially separated from the CMA for the auxiliary microphone-based SRP-PHAT spectra to be more reliable than the SRP-PHAT spectra without the auxiliary microphone. These findings are validated based on simulated microphone signals for several auxiliary microphone positions and two different noise and reverberation conditions., Comment: 5 pages, 3 figures, conference: EUSIPCO 2024 in Lyon

This paper describes the USTC-KXDIGIT system submitted to the ASVspoof5 Challenge for Track 1 (speech deepfake detection) and Track 2 (spoofing-robust automatic speaker verification, SASV). Track 1 showcases a diverse range of technical qualities from potential processing algorithms and includes both open and closed conditions. For these conditions, our system consists of a cascade of a frontend feature extractor and a back-end classifier. We focus on extensive embedding engineering and enhancing the generalization of the back-end classifier model. Specifically, the embedding engineering is based on hand-crafted features and speech representations from a self-supervised model, used for closed and open conditions, respectively. To detect spoof attacks under various adversarial conditions, we trained multiple systems on an augmented training set. Additionally, we used voice conversion technology to synthesize fake audio from genuine audio in the training set to enrich the synthesis algorithms. To leverage the complementary information learned by different model architectures, we employed activation ensemble and fused scores from different systems to obtain the final decision score for spoof detection. During the evaluation phase, the proposed methods achieved 0.3948 minDCF and 14.33% EER in the close condition, and 0.0750 minDCF and 2.59% EER in the open condition, demonstrating the robustness of our submitted systems under adversarial conditions. In Track 2, we continued using the CM system from Track 1 and fused it with a CNN-based ASV system. This approach achieved 0.2814 min-aDCF in the closed condition and 0.0756 min-aDCF in the open condition, showcasing superior performance in the SASV system., Comment: ASVspoof5 workshop paper

One-shot voice conversion(VC) aims to change the timbre of any source speech to match that of the target speaker with only one speech sample. Existing style transfer-based VC methods relied on speech representation disentanglement and suffered from accurately and independently encoding each speech component and recomposing back to converted speech effectively. To tackle this, we proposed Pureformer-VC, which utilizes Conformer blocks to build a disentangled encoder, and Zipformer blocks to build a style transfer decoder as the generator. In the decoder, we used effective styleformer blocks to integrate speaker characteristics effectively into the generated speech. The models used the generative VAE loss for encoding components and triplet loss for unsupervised discriminative training. We applied the styleformer method to Zipformer's shared weights for style transfer. The experimental results show that the proposed model achieves comparable subjective scores and exhibits improvements in objective metrics compared to existing methods in a one-shot voice conversion scenario., Comment: submmited to ICASSP 2025

Representing speech as discrete tokens provides a framework for transforming speech into a format that closely resembles text, thus enabling the use of speech as an input to the widely successful large language models (LLMs). Currently, while several speech tokenizers have been proposed, there is ambiguity regarding the properties that are desired from a tokenizer for specific downstream tasks and its overall generalizability. Evaluating the performance of tokenizers across different downstream tasks is a computationally intensive effort that poses challenges for scalability. To circumvent this requirement, we present STAB (Speech Tokenizer Assessment Benchmark), a systematic evaluation framework designed to assess speech tokenizers comprehensively and shed light on their inherent characteristics. This framework provides a deeper understanding of the underlying mechanisms of speech tokenization, thereby offering a valuable resource for expediting the advancement of future tokenizer models and enabling comparative analysis using a standardized benchmark. We evaluate the STAB metrics and correlate this with downstream task performance across a range of speech tasks and tokenizer choices., Comment: 5 pages

This work details our approach to achieving a leading system with a 1.79% pooled equal error rate (EER) on the evaluation set of the Controlled Singing Voice Deepfake Detection (CtrSVDD). The rapid advancement of generative AI models presents significant challenges for detecting AI-generated deepfake singing voices, attracting increased research attention. The Singing Voice Deepfake Detection (SVDD) Challenge 2024 aims to address this complex task. In this work, we explore the ensemble methods, utilizing speech foundation models to develop robust singing voice anti-spoofing systems. We also introduce a novel Squeeze-and-Excitation Aggregation (SEA) method, which efficiently and effectively integrates representation features from the speech foundation models, surpassing the performance of our other individual systems. Evaluation results confirm the efficacy of our approach in detecting deepfake singing voices. The codes can be accessed at https://github.com/Anmol2059/SVDD2024., Comment: Accepted to the IEEE Spoken Language Technology Workshop (SLT) 2024. Copyright may be transferred without notice, after which this version may no longer be accessible

In this paper, we propose long short term memory speech enhancement network (LSTMSE-Net), an audio-visual speech enhancement (AVSE) method. This innovative method leverages the complementary nature of visual and audio information to boost the quality of speech signals. Visual features are extracted with VisualFeatNet (VFN), and audio features are processed through an encoder and decoder. The system scales and concatenates visual and audio features, then processes them through a separator network for optimized speech enhancement. The architecture highlights advancements in leveraging multi-modal data and interpolation techniques for robust AVSE challenge systems. The performance of LSTMSE-Net surpasses that of the baseline model from the COG-MHEAR AVSE Challenge 2024 by a margin of 0.06 in scale-invariant signal-to-distortion ratio (SISDR), $0.03$ in short-time objective intelligibility (STOI), and $1.32$ in perceptual evaluation of speech quality (PESQ). The source code of the proposed LSTMSE-Net is available at \url{https://github.com/mtanveer1/AVSEC-3-Challenge}.

Diffusion-based voice conversion (VC) techniques such as VoiceGrad have attracted interest because of their high VC performance in terms of speech quality and speaker similarity. However, a notable limitation is the slow inference caused by the multi-step reverse diffusion. Therefore, we propose FastVoiceGrad, a novel one-step diffusion-based VC that reduces the number of iterations from dozens to one while inheriting the high VC performance of the multi-step diffusion-based VC. We obtain the model using adversarial conditional diffusion distillation (ACDD), leveraging the ability of generative adversarial networks and diffusion models while reconsidering the initial states in sampling. Evaluations of one-shot any-to-any VC demonstrate that FastVoiceGrad achieves VC performance superior to or comparable to that of previous multi-step diffusion-based VC while enhancing the inference speed. Audio samples are available at https://www.kecl.ntt.co.jp/people/kaneko.takuhiro/projects/fastvoicegrad/., Comment: Accepted to Interspeech 2024. Project page: https://www.kecl.ntt.co.jp/people/kaneko.takuhiro/projects/fastvoicegrad/

Transferring linguistic knowledge from a pretrained language model (PLM) to an acoustic model has been shown to greatly improve the performance of automatic speech recognition (ASR). However, due to the heterogeneous feature distributions in cross-modalities, designing an effective model for feature alignment and knowledge transfer between linguistic and acoustic sequences remains a challenging task. Optimal transport (OT), which efficiently measures probability distribution discrepancies, holds great potential for aligning and transferring knowledge between acoustic and linguistic modalities. Nonetheless, the original OT treats acoustic and linguistic feature sequences as two unordered sets in alignment and neglects temporal order information during OT coupling estimation. Consequently, a time-consuming pretraining stage is required to learn a good alignment between the acoustic and linguistic representations. In this paper, we propose a Temporal Order Preserved OT (TOT)-based Cross-modal Alignment and Knowledge Transfer (CAKT) (TOT-CAKT) for ASR. In the TOT-CAKT, local neighboring frames of acoustic sequences are smoothly mapped to neighboring regions of linguistic sequences, preserving their temporal order relationship in feature alignment and matching. With the TOT-CAKT model framework, we conduct Mandarin ASR experiments with a pretrained Chinese PLM for linguistic knowledge transfer. Our results demonstrate that the proposed TOT-CAKT significantly improves ASR performance compared to several state-of-the-art models employing linguistic knowledge transfer, and addresses the weaknesses of the original OT-based method in sequential feature alignment for ASR., Comment: Accepted to IEEE SLT 2024

Because videos in the wild can be out of sync for various reasons, a sync-net is used to bring the video back into sync for tasks that require synchronized videos. Previous state-of-the-art (SOTA) sync-nets use InfoNCE loss, rely on the transformer architecture, or both. Unfortunately, the former makes the model's output difficult to interpret, and the latter is unfriendly with large images, thus limiting the usefulness of sync-nets. In this work, we train a convolutional sync-net using the balanced BCE loss (BBCE), a loss inspired by the binary cross entropy (BCE) and the InfoNCE losses. In contrast to the InfoNCE loss, the BBCE loss does not require complicated sampling schemes. Our model can better handle larger images, and its output can be given a probabilistic interpretation. The probabilistic interpretation allows us to define metrics such as probability at offset and offscreen ratio to evaluate the sync quality of audio-visual (AV) speech datasets. Furthermore, our model achieves SOTA accuracy of $96.5\%$ on the LRS2 dataset and $93.8\%$ on the LRS3 dataset., Comment: 8+5 pages

In this paper, we introduce ConversaSynth, a framework designed to generate synthetic conversation audio using large language models (LLMs) with multiple persona settings. The framework first creates diverse and coherent text-based dialogues across various topics, which are then converted into audio using text-to-speech (TTS) systems. Our experiments demonstrate that ConversaSynth effectively generates highquality synthetic audio datasets, which can significantly enhance the training and evaluation of models for audio tagging, audio classification, and multi-speaker speech recognition. The results indicate that the synthetic datasets generated by ConversaSynth exhibit substantial diversity and realism, making them suitable for developing robust, adaptable audio-based AI systems., Comment: This work has been submitted for consideration at the WI-IAT'24 to be held in December 2024

The long speech sequence has been troubling language models (LM) based TTS approaches in terms of modeling complexity and efficiency. This work proposes SoCodec, a semantic-ordered multi-stream speech codec, to address this issue. It compresses speech into a shorter, multi-stream discrete semantic sequence with multiple tokens at each frame. Meanwhile, the ordered product quantization is proposed to constrain this sequence into an ordered representation. It can be applied with a multi-stream delayed LM to achieve better autoregressive generation along both time and stream axes in TTS. The experimental result strongly demonstrates the effectiveness of the proposed approach, achieving superior performance over baseline systems even if compressing the frameshift of speech from 20ms to 240ms (12x). The ablation studies further validate the importance of learning the proposed ordered multi-stream semantic representation in pursuing shorter speech sequences for efficient LM-based TTS., Comment: Accepted by SLT 2024

This paper introduces VoxHakka, a text-to-speech (TTS) system designed for Taiwanese Hakka, a critically under-resourced language spoken in Taiwan. Leveraging the YourTTS framework, VoxHakka achieves high naturalness and accuracy and low real-time factor in speech synthesis while supporting six distinct Hakka dialects. This is achieved by training the model with dialect-specific data, allowing for the generation of speaker-aware Hakka speech. To address the scarcity of publicly available Hakka speech corpora, we employed a cost-effective approach utilizing a web scraping pipeline coupled with automatic speech recognition (ASR)-based data cleaning techniques. This process ensured the acquisition of a high-quality, multi-speaker, multi-dialect dataset suitable for TTS training. Subjective listening tests conducted using comparative mean opinion scores (CMOS) demonstrate that VoxHakka significantly outperforms existing publicly available Hakka TTS systems in terms of pronunciation accuracy, tone correctness, and overall naturalness. This work represents a significant advancement in Hakka language technology and provides a valuable resource for language preservation and revitalization efforts., Comment: Submitted to O-COCOSDA 2024

Cross-domain speech enhancement (SE) is often faced with severe challenges due to the scarcity of noise and background information in an unseen target domain, leading to a mismatch between training and test conditions. This study puts forward a novel data simulation method to address this issue, leveraging noise-extractive techniques and generative adversarial networks (GANs) with only limited target noisy speech data. Notably, our method employs a noise encoder to extract noise embeddings from target-domain data. These embeddings aptly guide the generator to synthesize utterances acoustically fitted to the target domain while authentically preserving the phonetic content of the input clean speech. Furthermore, we introduce the notion of dynamic stochastic perturbation, which can inject controlled perturbations into the noise embeddings during inference, thereby enabling the model to generalize well to unseen noise conditions. Experiments on the VoiceBank-DEMAND benchmark dataset demonstrate that our domain-adaptive SE method outperforms an existing strong baseline based on data simulation., Comment: Accepted to IEEE SLT 2024

Speech foundation models have achieved state-of-the-art (SoTA) performance across various tasks, such as automatic speech recognition (ASR) in hundreds of languages. However, multi-speaker ASR remains a challenging task for these models due to data scarcity and sparsity. In this paper, we present approaches to enable speech foundation models to process and understand multi-speaker speech with limited training data. Specifically, we adapt a speech foundation model for the multi-speaker ASR task using only telephonic data. Remarkably, the adapted model also performs well on meeting data without any fine-tuning, demonstrating the generalization ability of our approach. We conduct several ablation studies to analyze the impact of different parameters and strategies on model performance. Our findings highlight the effectiveness of our methods. Results show that less parameters give better overall cpWER, which, although counter-intuitive, provides insights into adapting speech foundation models for multi-speaker ASR tasks with minimal annotated data., Comment: Accepted by SLT 2024

Recently, attention-based transformers have become a de facto standard in many deep learning applications including natural language processing, computer vision, signal processing, etc.. In this paper, we propose a transformer-based end-to-end model to extract a target speaker's speech from a monaural multi-speaker mixed audio signal. Unlike existing speaker extraction methods, we introduce two additional objectives to impose speaker embedding consistency and waveform encoder invertibility and jointly train both speaker encoder and speech separator to better capture the speaker conditional embedding. Furthermore, we leverage a multi-scale discriminator to refine the perceptual quality of the extracted speech. Our experiments show that the use of a dual path transformer in the separator backbone along with proposed training paradigm improves the CNN baseline by $3.12$ dB points. Finally, we compare our approach with recent state-of-the-arts and show that our model outperforms existing methods by $4.1$ dB points on an average without creating additional data dependency.

Recent speech technologies have led to produce high quality synthesised speech due to recent advances in neural Text to Speech (TTS). However, such TTS models depend on extensive amounts of data that can be costly to produce and is hardly scalable to all existing languages, especially that seldom attention is given to low resource languages. With techniques such as knowledge transfer, the burden of creating datasets can be alleviated. In this paper, we therefore investigate two aspects; firstly, whether data from social media can be used for a small TTS dataset construction, and secondly whether cross lingual transfer learning (TL) for a low resource language can work with this type of data. In this aspect, we specifically assess to what extent multilingual modeling can be leveraged as an alternative to training on monolingual corporas. To do so, we explore how data from foreign languages may be selected and pooled to train a TTS model for a target low resource language. Our findings show that multilingual pre-training is better than monolingual pre-training at increasing the intelligibility and naturalness of the generated speech., Comment: 12 pages, 2 figures

Although automatic pathological speech detection approaches show promising results when clean recordings are available, they are vulnerable to additive noise. Recently it has been shown that databases commonly used to develop and evaluate such approaches are noisy, with the noise characteristics between healthy and pathological recordings being different. Consequently, automatic approaches trained on these databases often learn to discriminate noise rather than speech pathology. This paper introduces a method to mitigate this noise disparity in training data. Using noise estimates from recordings from one group of speakers to augment recordings from the other group, the noise characteristics become consistent across all recordings. Experimental results demonstrate the efficacy of this approach in mitigating noise disparity in training data, thereby enabling automatic pathological speech detection to focus on pathology-discriminant cues rather than noise-discriminant ones., Comment: To appear in IWAENC 2024

In this work, we aim to analyze and optimize the EnCLAP framework, a state-of-the-art model in automated audio captioning. We investigate the impact of modifying the acoustic encoder components, explore pretraining with different dataset scales, and study the effectiveness of a reranking scheme. Through extensive experimentation and quantitative analysis of generated captions, we develop EnCLAP++, an enhanced version that significantly surpasses the original., Comment: Accepted to DCASE2024 Workshop

In this technical report, we describe our submission to DCASE2024 Challenge Task6 (Automated Audio Captioning) and Task8 (Language-based Audio Retrieval). We develop our approach building upon the EnCLAP audio captioning framework and optimizing it for Task6 of the challenge. Notably, we outline the changes in the underlying components and the incorporation of the reranking process. Additionally, we submit a supplementary retriever model, a byproduct of our modified framework, to Task8. Our proposed systems achieve FENSE score of 0.542 on Task6 and mAP@10 score of 0.386 on Task8, significantly outperforming the baseline models., Comment: DCASE2024 Challenge Technical Report. Ranked 2nd in Task 6 Automated Audio Captioning

We propose a novel symbolic music representation and Generative Adversarial Network (GAN) framework specially designed for symbolic multitrack music generation. The main theme of symbolic music generation primarily encompasses the preprocessing of music data and the implementation of a deep learning framework. Current techniques dedicated to symbolic music generation generally encounter two significant challenges: training data's lack of information about chords and scales and the requirement of specially designed model architecture adapted to the unique format of symbolic music representation. In this paper, we solve the above problems by introducing new symbolic music representation with MusicLang chord analysis model. We propose our MMT-BERT architecture adapting to the representation. To build a robust multitrack music generator, we fine-tune a pre-trained MusicBERT model to serve as the discriminator, and incorporate relativistic standard loss. This approach, supported by the in-depth understanding of symbolic music encoded within MusicBERT, fortifies the consonance and humanity of music generated by our method. Experimental results demonstrate the effectiveness of our approach which strictly follows the state-of-the-art methods., Comment: Accepted to the 25th International Society for Music Information Retrieval Conference (ISMIR 2024)

Recent advancements in machine learning have fueled research on multimodal tasks, such as for instance text-to-video and text-to-audio retrieval. These tasks require models to understand the semantic content of video and audio data, including objects, and characters. The models also need to learn spatial arrangements and temporal relationships. In this work, we analyse the temporal ordering of sounds, which is an understudied problem in the context of text-to-audio retrieval. In particular, we dissect the temporal understanding capabilities of a state-of-the-art model for text-to-audio retrieval on the AudioCaps and Clotho datasets. Additionally, we introduce a synthetic text-audio dataset that provides a controlled setting for evaluating temporal capabilities of recent models. Lastly, we present a loss function that encourages text-audio models to focus on the temporal ordering of events. Code and data are available at https://www.robots.ox.ac.uk/~vgg/research/audio-retrieval/dtu/., Comment: 9 pages, 5 figures, ACM Multimedia 2024, https://www.robots.ox.ac.uk/~vgg/research/audio-retrieval/dtu/