Existing methods for synthesizing 3D human gestures from speech have shown promising results, but they do not explicitly model the impact of emotions on the generated gestures. Instead, these methods directly output animations from speech without control over the expressed emotion. To address this limitation, we present AMUSE, an emotional speech-driven body animation model based on latent diffusion. Our observation is that content (i.e., gestures related to speech rhythm and word utterances), emotion, and personal style are separable. To account for this, AMUSE maps the driving audio to three disentangled latent vectors: one for content, one for emotion, and one for personal style. A latent diffusion model, trained to generate gesture motion sequences, is then conditioned on these latent vectors. Once trained, AMUSE synthesizes 3D human gestures directly from speech with control over the expressed emotions and style by combining the content from the driving speech with the emotion and style of another speech sequence. Randomly sampling the noise of the diffusion model further generates variations of the gesture with the same emotional expressivity. Qualitative, quantitative, and perceptual evaluations demonstrate that AMUSE outputs realistic gesture sequences. Compared to the state of the art, the generated gestures are better synchronized with the speech content and better represent the emotion expressed by the input speech. Our project website is amuse.is.tue.mpg.de.
Various new brain-computer interface technologies or neuroscience applications require decoding stimulus-following neural responses to natural stimuli such as speech and video from, e.g., electroencephalography (EEG) signals. In this context, generalized canonical correlation analysis (GCCA) is often used as a group analysis technique, which allows the extraction of correlated signal components from the neural activity of multiple subjects attending to the same stimulus. GCCA can be used to improve the signal-to-noise ratio of the stimulus-following neural responses relative to all other irrelevant (non-)neural activity, or to quantify the correlated neural activity across multiple subjects in a group-wise coherence metric. However, the traditional GCCA technique is stimulus-unaware: no information about the stimulus is used to estimate the correlated components from the neural data of several subjects. Therefore, the GCCA technique might fail to extract relevant correlated signal components in practical situations where the amount of information is limited, for example, because of a limited amount of training data or group size. This motivates a new stimulus-informed GCCA (SI-GCCA) framework that allows taking the stimulus into account to extract the correlated components. We show that SI-GCCA outperforms GCCA in various practical settings, for both auditory and visual stimuli. Moreover, we showcase how SI-GCCA can be used to steer the estimation of the components towards the stimulus. As such, SI-GCCA substantially improves upon GCCA for various purposes, ranging from preprocessing to quantifying attention.
Prompt-based methods have been successfully applied to multilingual pretrained language models for zero-shot cross-lingual understanding. However, most previous studies primarily focused on sentence-level classification tasks, and only a few considered token-level labeling tasks such as Named Entity Recognition (NER) and Part-of-Speech (POS) tagging. In this paper, we propose Token-Level Prompt Decomposition (ToPro), which facilitates the prompt-based method for token-level sequence labeling tasks. The ToPro method decomposes an input sentence into single tokens and applies one prompt template to each token. Our experiments on multilingual NER and POS tagging datasets demonstrate that ToPro-based fine-tuning outperforms Vanilla fine-tuning and Prompt-Tuning in zero-shot cross-lingual transfer, especially for languages that are typologically different from the source language English. Our method also attains state-of-the-art performance when employed with the mT5 model. Besides, our exploratory study in multilingual large language models shows that ToPro performs much better than the current in-context learning method. Overall, the performance improvements show that ToPro could potentially serve as a novel and simple benchmarking method for sequence labeling tasks.
Challenges exist in learning and understanding religions, such as the complexity and depth of religious doctrines and teachings. Chatbots as question-answering systems can help in solving these challenges. LLM chatbots use NLP techniques to establish connections between topics and accurately respond to complex questions. These capabilities make it perfect for enlightenment on religion as a question-answering chatbot. However, LLMs also tend to generate false information, known as hallucination. Also, the chatbots' responses can include content that insults personal religious beliefs, interfaith conflicts, and controversial or sensitive topics. It must avoid such cases without promoting hate speech or offending certain groups of people or their beliefs. This study uses a vector database-based Retrieval Augmented Generation (RAG) approach to enhance the accuracy and transparency of LLMs. Our question-answering system is called "MufassirQAS". We created a database consisting of several open-access books that include Turkish context. These books contain Turkish translations and interpretations of Islam. This database is utilized to answer religion-related questions and ensure our answers are trustworthy. The relevant part of the dataset, which LLM also uses, is presented along with the answer. We have put careful effort into creating system prompts that give instructions to prevent harmful, offensive, or disrespectful responses to respect people's values and provide reliable results. The system answers and shares additional information, such as the page number from the respective book and the articles referenced for obtaining the information. MufassirQAS and ChatGPT are also tested with sensitive questions. We got better performance with our system. Study and enhancements are still in progress. Results and future works are given.
Large Language Models (LLMs) have demonstrated superior abilities in tasks such as chatting, reasoning, and question-answering. However, standard LLMs may ignore crucial paralinguistic information, such as sentiment, emotion, and speaking style, which are essential for achieving natural, human-like spoken conversation, especially when such information is conveyed by acoustic cues. We therefore propose Paralinguistics-enhanced Generative Pretrained Transformer (ParalinGPT), an LLM that utilizes text and speech modalities to better model the linguistic content and paralinguistic attributes of spoken dialogue. The model takes the conversational context of text, speech embeddings, and paralinguistic attributes as input prompts within a serialized multitasking multimodal framework. Specifically, our framework serializes tasks in the order of current paralinguistic attribute prediction, response paralinguistic attribute prediction, and response text generation with autoregressive conditioning. We utilize the Switchboard-1 corpus, including its sentiment labels as the paralinguistic attribute, as our spoken dialogue dataset. Experimental results indicate the proposed serialized multitasking method outperforms typical sequence classification techniques on current and response sentiment classification. Furthermore, leveraging conversational context and speech embeddings significantly improves both response text generation and sentiment prediction. Our proposed framework achieves relative improvements of 6.7%, 12.0%, and 3.5% in current sentiment accuracy, response sentiment accuracy, and response text BLEU score, respectively.
Diffusion models are a new class of generative models that have shown outstanding performance in image generation literature. As a consequence, studies have attempted to apply diffusion models to other tasks, such as speech enhancement. A popular approach in adapting diffusion models to speech enhancement consists in modelling a progressive transformation between the clean and noisy speech signals. However, one popular diffusion model framework previously laid in image generation literature did not account for such a transformation towards the system input, which prevents from relating the existing diffusion-based speech enhancement systems with the aforementioned diffusion model framework. To address this, we extend this framework to account for the progressive transformation between the clean and noisy speech signals. This allows us to apply recent developments from image generation literature, and to systematically investigate design aspects of diffusion models that remain largely unexplored for speech enhancement, such as the neural network preconditioning, the training loss weighting, the stochastic differential equation (SDE), or the amount of stochasticity injected in the reverse process. We show that the performance of previous diffusion-based speech enhancement systems cannot be attributed to the progressive transformation between the clean and noisy speech signals. Moreover, we show that a proper choice of preconditioning, training loss weighting, SDE and sampler allows to outperform a popular diffusion-based speech enhancement system in terms of perceptual metrics while using fewer sampling steps, thus reducing the computational cost by a factor of four.
This paper describes our homophobia/transphobia in social media comments detection system developed as part of the shared task at LT-EDI-2024. We took a transformer-based approach to develop our multiclass classification model for ten language conditions (English, Spanish, Gujarati, Hindi, Kannada, Malayalam, Marathi, Tamil, Tulu, and Telugu). We introduced synthetic and organic instances of script-switched language data during domain adaptation to mirror the linguistic realities of social media language as seen in the labelled training data. Our system ranked second for Gujarati and Telugu with varying levels of performance for other language conditions. The results suggest incorporating elements of paralinguistic behaviour such as script-switching may improve the performance of language detection systems especially in the cases of under-resourced languages conditions.
The style transfer task in Text-to-Speech refers to the process of transferring style information into text content to generate corresponding speech with a specific style. However, most existing style transfer approaches are either based on fixed emotional labels or reference speech clips, which cannot achieve flexible style transfer. Recently, some methods have adopted text descriptions to guide style transfer. In this paper, we propose a more flexible multi-modal and style controllable TTS framework named MM-TTS. It can utilize any modality as the prompt in unified multi-modal prompt space, including reference speech, emotional facial images, and text descriptions, to control the style of the generated speech in a system. The challenges of modeling such a multi-modal style controllable TTS mainly lie in two aspects:1)aligning the multi-modal information into a unified style space to enable the input of arbitrary modality as the style prompt in a single system, and 2)efficiently transferring the unified style representation into the given text content, thereby empowering the ability to generate prompt style-related voice. To address these problems, we propose an aligned multi-modal prompt encoder that embeds different modalities into a unified style space, supporting style transfer for different modalities. Additionally, we present a new adaptive style transfer method named Style Adaptive Convolutions to achieve a better style representation. Furthermore, we design a Rectified Flow based Refiner to solve the problem of over-smoothing Mel-spectrogram and generate audio of higher fidelity. Since there is no public dataset for multi-modal TTS, we construct a dataset named MEAD-TTS, which is related to the field of expressive talking head. Our experiments on the MEAD-TTS dataset and out-of-domain datasets demonstrate that MM-TTS can achieve satisfactory results based on multi-modal prompts.
Several sensing techniques have been proposed for silent speech recognition (SSR); however, many of these methods require invasive processes or sensor attachment to the skin using adhesive tape or glue, rendering them unsuitable for frequent use in daily life. By contrast, impulse radio ultra-wideband (IR-UWB) radar can operate without physical contact with users' articulators and related body parts, offering several advantages for SSR. These advantages include high range resolution, high penetrability, low power consumption, robustness to external light or sound interference, and the ability to be embedded in space-constrained handheld devices. This study demonstrated IR-UWB radar-based contactless SSR using four types of speech stimuli (vowels, consonants, words, and phrases). To achieve this, a novel speech feature extraction algorithm specifically designed for IR-UWB radar-based SSR is proposed. Each speech stimulus is recognized by applying a classification algorithm to the extracted speech features. Two different algorithms, multidimensional dynamic time warping (MD-DTW) and deep neural network-hidden Markov model (DNN-HMM), were compared for the classification task. Additionally, a favorable radar antenna position, either in front of the user's lips or below the user's chin, was determined to achieve higher recognition accuracy. Experimental results demonstrated the efficacy of the proposed speech feature extraction algorithm combined with DNN-HMM for classifying vowels, consonants, words, and phrases. Notably, this study represents the first demonstration of phoneme-level SSR using contactless radar.
Emotion recognition (ER) from speech signals is a robust approach since it cannot be imitated like facial expression or text based sentiment analysis. Valuable information underlying the emotions are significant for human-computer interactions enabling intelligent machines to interact with sensitivity in the real world. Previous ER studies through speech signal processing have focused exclusively on associations between different signal mode decomposition methods and hidden informative features. However, improper decomposition parameter selections lead to informative signal component losses due to mode duplicating and mixing. In contrast, the current study proposes VGG-optiVMD, an empowered variational mode decomposition algorithm, to distinguish meaningful speech features and automatically select the number of decomposed modes and optimum balancing parameter for the data fidelity constraint by assessing their effects on the VGG16 flattening output layer. Various feature vectors were employed to train the VGG16 network on different databases and assess VGG-optiVMD reproducibility and reliability. One, two, and three-dimensional feature vectors were constructed by concatenating Mel-frequency cepstral coefficients, Chromagram, Mel spectrograms, Tonnetz diagrams, and spectral centroids. Results confirmed a synergistic relationship between the fine-tuning of the signal sample rate and decomposition parameters with classification accuracy, achieving state-of-the-art 96.09% accuracy in predicting seven emotions on the Berlin EMO-DB database.