This paper introduces UnDiff, a diffusion probabilistic model capable of solving various speech inverse tasks. Being once trained for speech waveform generation in an unconditional manner, it can be adapted to different tasks including degradation inversion, neural vocoding, and source separation. In this paper, we, first, tackle the challenging problem of unconditional waveform generation by comparing different neural architectures and preconditioning domains. After that, we demonstrate how the trained unconditional diffusion could be adapted to different tasks of speech processing by the means of recent developments in post-training conditioning of diffusion models. Finally, we demonstrate the performance of the proposed technique on the tasks of bandwidth extension, declipping, vocoding, and speech source separation and compare it to the baselines. The codes will be released soon.
In this study, we address the importance of modeling behavior style in virtual agents for personalized human-agent interaction. We propose a machine learning approach to synthesize gestures, driven by prosodic features and text, in the style of different speakers, even those unseen during training. Our model incorporates zero-shot multimodal style transfer using multimodal data from the PATS database, which contains videos of diverse speakers. We recognize style as a pervasive element during speech, influencing the expressivity of communicative behaviors, while content is conveyed through multimodal signals and text. By disentangling content and style, we directly infer the style embedding, even for speakers not included in the training phase, without the need for additional training or fine-tuning. Objective and subjective evaluations are conducted to validate our approach and compare it against two baseline methods.
The emotion detection technology to enhance human decision-making is an important research issue for real-world applications, but real-life emotion datasets are relatively rare and small. The experiments conducted in this paper use the CEMO, which was collected in a French emergency call center. Two pre-trained models based on speech and text were fine-tuned for speech emotion recognition. Using pre-trained Transformer encoders mitigates our data's limited and sparse nature. This paper explores the different fusion strategies of these modality-specific models. In particular, fusions with and without cross-attention mechanisms were tested to gather the most relevant information from both the speech and text encoders. We show that multimodal fusion brings an absolute gain of 4-9% with respect to either single modality and that the Symmetric multi-headed cross-attention mechanism performed better than late classical fusion approaches. Our experiments also suggest that for the real-life CEMO corpus, the audio component encodes more emotive information than the textual one.
We study speech enhancement using deep learning (DL) for virtual meetings on cellular devices, where transmitted speech has background noise and transmission loss that affects speech quality. Since the Deep Noise Suppression (DNS) Challenge dataset does not contain practical disturbance, we collect a transmitted DNS (t-DNS) dataset using Zoom Meetings over T-Mobile network. We select two baseline models: Demucs and FullSubNet. The Demucs is an end-to-end model that takes time-domain inputs and outputs time-domain denoised speech, and the FullSubNet takes time-frequency-domain inputs and outputs the energy ratio of the target speech in the inputs. The goal of this project is to enhance the speech transmitted over the cellular networks using deep learning models.
The goal of this work is zero-shot text-to-speech synthesis, with speaking styles and voices learnt from facial characteristics. Inspired by the natural fact that people can imagine the voice of someone when they look at his or her face, we introduce a face-styled diffusion text-to-speech (TTS) model within a unified framework learnt from visible attributes, called Face-TTS. This is the first time that face images are used as a condition to train a TTS model. We jointly train cross-model biometrics and TTS models to preserve speaker identity between face images and generated speech segments. We also propose a speaker feature binding loss to enforce the similarity of the generated and the ground truth speech segments in speaker embedding space. Since the biometric information is extracted directly from the face image, our method does not require extra fine-tuning steps to generate speech from unseen and unheard speakers. We train and evaluate the model on the LRS3 dataset, an in-the-wild audio-visual corpus containing background noise and diverse speaking styles. The project page is https://facetts.github.io.
Verifying the integrity of voice recording evidence for criminal investigations is an integral part of an audio forensic analyst's work. Here, one focus is on detecting deletion or insertion operations, so called audio splicing. While this is a rather easy approach to alter spoken statements, careful editing can yield quite convincing results. For difficult cases or big amounts of data, automated tools can support in detecting potential editing locations. To this end, several analytical and deep learning methods have been proposed by now. Still, few address unconstrained splicing scenarios as expected in practice. With SigPointer, we propose a pointer network framework for continuous input that uncovers splice locations naturally and more efficiently than existing works. Extensive experiments on forensically challenging data like strongly compressed and noisy signals quantify the benefit of the pointer mechanism with performance increases between about 6 to 10 percentage points.
While human evaluation is the most reliable metric for evaluating speech generation systems, it is generally costly and time-consuming. Previous studies on automatic speech quality assessment address the problem by predicting human evaluation scores with machine learning models. However, they rely on supervised learning and thus suffer from high annotation costs and domain-shift problems. We propose SpeechLMScore, an unsupervised metric to evaluate generated speech using a speech-language model. SpeechLMScore computes the average log-probability of a speech signal by mapping it into discrete tokens and measures the average probability of generating the sequence of tokens. Therefore, it does not require human annotation and is a highly scalable framework. Evaluation results demonstrate that the proposed metric shows a promising correlation with human evaluation scores on different speech generation tasks including voice conversion, text-to-speech, and speech enhancement.
While recent research has made significant progress in speech-driven talking face generation, the quality of the generated video still lags behind that of real recordings. One reason for this is the use of handcrafted intermediate representations like facial landmarks and 3DMM coefficients, which are designed based on human knowledge and are insufficient to precisely describe facial movements. Additionally, these methods require an external pretrained model for extracting these representations, whose performance sets an upper bound on talking face generation. To address these limitations, we propose a novel method called DAE-Talker that leverages data-driven latent representations obtained from a diffusion autoencoder (DAE). DAE contains an image encoder that encodes an image into a latent vector and a DDIM image decoder that reconstructs the image from it. We train our DAE on talking face video frames and then extract their latent representations as the training target for a Conformer-based speech2latent model. This allows DAE-Talker to synthesize full video frames and produce natural head movements that align with the content of speech, rather than relying on a predetermined head pose from a template video. We also introduce pose modelling in speech2latent for pose controllability. Additionally, we propose a novel method for generating continuous video frames with the DDIM image decoder trained on individual frames, eliminating the need for modelling the joint distribution of consecutive frames directly. Our experiments show that DAE-Talker outperforms existing popular methods in lip-sync, video fidelity, and pose naturalness. We also conduct ablation studies to analyze the effectiveness of the proposed techniques and demonstrate the pose controllability of DAE-Talker.
Language development experts need tools that can automatically identify languages from fluent, conversational speech, and provide reliable estimates of usage rates at the level of an individual recording. However, language identification systems are typically evaluated on metrics such as equal error rate and balanced accuracy, applied at the level of an entire speech corpus. These overview metrics do not provide information about model performance at the level of individual speakers, recordings, or units of speech with different linguistic characteristics. Overview statistics may therefore mask systematic errors in model performance for some subsets of the data, and consequently, have worse performance on data derived from some subsets of human speakers, creating a kind of algorithmic bias. In the current paper, we investigate how well a number of language identification systems perform on individual recordings and speech units with different linguistic properties in the MERLIon CCS Challenge. The Challenge dataset features accented English-Mandarin code-switched child-directed speech.
Most lip-to-speech (LTS) synthesis models are trained and evaluated under the assumption that the audio-video pairs in the dataset are perfectly synchronized. In this work, we show that the commonly used audio-visual datasets, such as GRID, TCD-TIMIT, and Lip2Wav, can have data asynchrony issues. Training lip-to-speech with such datasets may further cause the model asynchrony issue -- that is, the generated speech and the input video are out of sync. To address these asynchrony issues, we propose a synchronized lip-to-speech (SLTS) model with an automatic synchronization mechanism (ASM) to correct data asynchrony and penalize model asynchrony. We further demonstrate the limitation of the commonly adopted evaluation metrics for LTS with asynchronous test data and introduce an audio alignment frontend before the metrics sensitive to time alignment for better evaluation. We compare our method with state-of-the-art approaches on conventional and time-aligned metrics to show the benefits of synchronization training.