is professor of Health Informatics at TU Munich.
His research combines computer science with modern health care and medicine. The main focus lies in the acquisition, analysis, and interpretation of biosignals including in daily life, such as those generated in monitoring heart activity, metabolism, or neuronal activities. Additionally, acoustic, visual, and a variety of other parameters are also evaluated. The goal is prevention, diagnosis, as well as decision support and intervention through efficient, transparent, and trustworthy methods of current Artificial Intelligence.
Increasingly frequent publications in the literature report voice quality differences between depressed patients and controls. Here, we examine the possibility of using voice analysis as an early warning signal for the development of emotion disturbances in young adults. As part of a major interdisciplinary European research project in four countries (ECoWeB), examining the effects of web-based prevention programs to reduce the risk for depression in young adults, we analyzed a large number of acoustic voice characteristics in vocal reports of emotions experienced by the participants on a specific day. We were able to identify a number of significant differences in acoustic cues, particularly with respect to the energy distribution in the voice spectrum, encouraging further research efforts to develop promising non-obtrusive risk indicators in the normal speaking voice. This is particularly important in the case of young adults who are less likely to exhibit standard risk factors for depression such as negative life experiences.
Curriculum learning (CL) describes a machine learning training strategy in which samples are gradually introduced into the training process based on their difficulty. Despite a partially contradictory body of evidence in the literature, CL finds popularity in deep learning research due to its promise of leveraging human-inspired curricula to achieve higher model performance. Yet, the subjectivity and biases that follow any necessary definition of difficulty, especially for those found in orderings derived from models or training statistics, have rarely been investigated. To shed more light on the underlying unanswered questions, we conduct an extensive study on the robustness and similarity of the most common scoring functions for sample difficulty estimation, as well as their potential benefits in CL, using the popular benchmark dataset CIFAR-10 and the acoustic scene classification task from the DCASE2020 challenge as representatives of computer vision and computer audition, respectively. We report a strong dependence of scoring functions on the training setting, including randomness, which can partly be mitigated through ensemble scoring. While we do not find a general advantage of CL over uniform sampling, we observe that the ordering in which data is presented for CL-based training plays an important role in model performance. Furthermore, we find that the robustness of scoring functions across random seeds positively correlates with CL performance. Finally, we uncover that models trained with different CL strategies complement each other by boosting predictive power through late fusion, likely due to differences in the learnt concepts. Alongside our findings, we release the aucurriculum toolkit (this https URL), implementing sample difficulty and CL-based training in a modular fashion.
Audio-based kinship verification (AKV) is important in many domains, such as home security monitoring, forensic identification, and social network analysis. A key challenge in the task arises from differences in age across samples from different individuals, which can be interpreted as a domain bias in a cross-domain verification task. To address this issue, we design the notion of an ‘age-standardised domain’ wherein we utilise the optimised CycleGAN-VC3 network to perform age-audio conversion to generate the in-domain audio. The generated audio dataset is employed to extract a range of features, which are then fed into a metric learning architecture to verify kinship. Experiments are conducted on the KAN_AV audio dataset, which contains age and kinship labels. The results demonstrate that the method markedly enhances the accuracy of kinship verification, while also offering novel insights for future kinship verification research.
Neural network models for audio tasks, such as automatic speech recognition (ASR) and acoustic scene classification (ASC), are susceptible to noise contamination for real-life applications. To improve audio quality, an enhancement module, which can be developed independently, is explicitly used at the front-end of the target audio applications. In this paper, we present an end-to-end learning solution to jointly optimise the models for audio enhancement (AE) and the subsequent applications. To guide the optimisation of the AE module towards a target application, and especially to overcome difficult samples, we make use of the sample-wise performance measure as an indication of sample importance. In experiments, we consider four representative applications to evaluate our training paradigm, i.e., ASR, speech command recognition (SCR), speech emotion recognition (SER), and ASC. These applications are associated with speech and nonspeech tasks concerning semantic and non-semantic features, transient and global information, and the experimental results indicate that our proposed approach can considerably boost the noise robustness of the models, especially at low signal-to-noise ratios, for a wide range of computer audition tasks in everyday-life noisy environments.
While current emotional text-to-speech (TTS) systems can generate highly intelligible emotional speech, achieving fine control over emotion rendering of the output speech still remains a significant challenge. In this paper, we introduce ParaEVITS, a novel emotional TTS framework that leverages the compositionality of natural language to enhance control over emotional rendering. By incorporating a text-audio encoder inspired by ParaCLAP, a contrastive language-audio pretraining (CLAP) model for computational paralinguistics, the diffusion model is trained to generate emotional embeddings based on textual emotional style descriptions. Our framework first trains on reference audio using the audio encoder, then fine-tunes a diffusion model to process textual inputs from ParaCLAP’s text encoder. During inference, speech attributes such as pitch, jitter, and loudness are manipulated using only textual conditioning. Our experiments demonstrate that ParaEVITS effectively control emotion rendering without compromising speech quality. Speech demos are publicly available.
Prior work in computational bioacoustics has mostly focused on the detection of animal presence in a particular habitat. However, animal sounds contain much richer information than mere presence; among others, they encapsulate the interactions of those animals with other members of their species. Studying these interactions is almost impossible in a naturalistic setting, as the ground truth is often lacking. The use of animals in captivity instead offers a viable alternative pathway. However, most prior works follow a traditional, statistics-based approach to analysing interactions. In the present work, we go beyond this standard framework by attempting to predict the underlying context in interactions between captive Rousettus Aegyptiacus using deep neural networks. We reach an unweighted average recall of over 30% - more than thrice the chance level - and show error patterns that differ from our statistical analysis. This work thus represents an important step towards the automatic analysis of states in animals from sound.
Telling stories is an integral part of human communication which can evoke emotions and influence the affective states of the audience. Automatically modeling emotional trajectories in stories has thus attracted considerable scholarly interest. However, as most existing works have been limited to unsupervised dictionary-based approaches, there is no benchmark for this task. We address this gap by introducing continuous valence and arousal labels for an existing dataset of children’s stories originally annotated with discrete emotion categories. We collect additional annotations for this data and map the categorical labels to the continuous valence and arousal space. For predicting the thus obtained emotionality signals, we fine-tune a DeBERTa model and improve upon this baseline via a weakly supervised learning approach. The best configuration achieves a Concordance Correlation Coefficient (CCC) of .8221 for valence and .7125 for arousal on the test set, demonstrating the efficacy of our proposed approach. A detailed analysis shows the extent to which the results vary depending on factors such as the author, the individual story, or the section within the story. In addition, we uncover the weaknesses of our approach by investigating examples that prove to be difficult to predict.
Speech Emotion Recognition (SER) is crucial for enabling computers to understand the emotions conveyed in human communication. With recent advancements in Deep Learning (DL), the performance of SER models has significantly improved. However, designing an optimal DL architecture requires specialised knowledge and experimental assessments. Fortunately, Neural Architecture Search (NAS) provides a potential solution for automatically determining the best DL model. The Differentiable Architecture Search (DARTS) is a particularly efficient method for discovering optimal models. This study presents emoDARTS, a DARTS-optimised joint CNN and Sequential Neural Network (SeqNN: LSTM, RNN) architecture that enhances SER performance. The literature supports the selection of CNN and LSTM coupling to improve performance. While DARTS has previously been used to choose CNN and LSTM operations independently, our technique adds a novel mechanism for selecting CNN and SeqNN operations in conjunction using DARTS. Unlike earlier work, we do not impose limits on the layer order of the CNN. Instead, we let DARTS choose the best layer order inside the DARTS cell. We demonstrate that emoDARTS outperforms conventionally designed CNN-LSTM models and surpasses the best-reported SER results achieved through DARTS on CNN-LSTM by evaluating our approach on the IEMOCAP, MSP-IMPROV, and MSP-Podcast datasets.
Ubiquitous sensing has been widely applied in smart healthcare, providing an opportunity for intelligent heart sound auscultation. However, smart devices contain sensitive information, raising user privacy concerns. To this end, federated learning (FL) has been adopted as an effective solution, enabling decentralised learning without data sharing, thus preserving data privacy in the Internet of Health Things (IoHT). Nevertheless, traditional FL requires the same architectural models to be trained across local clients and global servers, leading to a lack of model heterogeneity and client personalisation. For medical institutions with private data clients, this study proposes Fed-MStacking, a heterogeneous FL framework that incorporates a stacking ensemble learning strategy to support clients in building their own models. The secondary objective of this study is to address scenarios involving local clients with data characterised by inconsistent labelling. Specifically, the local client contains only one case type, and the data cannot be shared within or outside the institution. To train a global multi-class classifier, we aggregate missing class information from all clients at each institution and build meta-data, which then participates in FL training via a meta-learner. We apply the proposed framework to a multi-institutional heart sound database. The experiments utilise random forests (RFs), feedforward neural networks (FNNs), and convolutional neural networks (CNNs) as base classifiers. The results show that the heterogeneous stacking of local models performs better compared to homogeneous stacking.
In emergency medicine, timely intervention for patients at risk of suicide is often hindered by delayed access to specialised psychiatric care. To bridge this gap, we introduce a speech-based approach for automatic suicide risk assessment. Our study involves a novel dataset comprising speech recordings of 20 patients who read neutral texts. We extract four speech representations encompassing interpretable and deep features. Further, we explore the impact of gender-based modelling and phrase-level normalisation. By applying gender-exclusive modelling, features extracted from an emotion fine-tuned wav2vec2.0 model can be utilised to discriminate high- from low-suicide risk with a balanced accuracy of 81%. Finally, our analysis reveals a discrepancy in the relationship of speech characteristics and suicide risk between female and male subjects. For men in our dataset, suicide risk increases together with agitation while voice characteristics of female subjects point the other way.
Foundation models have shown great promise in speech emotion recognition (SER) by leveraging their pre-trained representations to capture emotion patterns in speech signals. To further enhance SER performance across various languages and domains, we propose a novel twofold approach. First, we gather EmoSet++, a comprehensive multi-lingual, multi-cultural speech emotion corpus with 37 datasets, 150,907 samples, and a total duration of 119.5 hours. Second, we introduce ExHuBERT, an enhanced version of HuBERT achieved by backbone extension and fine-tuning on EmoSet++. We duplicate each encoder layer and its weights, then freeze the first duplicate, integrating an extra zero-initialized linear layer and skip connections to preserve functionality and ensure its adaptability for subsequent fine-tuning. Our evaluation on unseen datasets shows the efficacy of ExHuBERT, setting a new benchmark for various SER tasks.
Flattery is an important aspect of human communication that facilitates social bonding, shapes perceptions, and influences behavior through strategic compliments and praise, leveraging the power of speech to build rapport effectively. Its automatic detection can thus enhance the naturalness of human-AI interactions. To meet this need, we present a novel audio textual dataset comprising 20 hours of speech and train machine learning models for automatic flattery detection. In particular, we employ pretrained AST, Wav2Vec2, and Whisper models for the speech modality, and Whisper TTS models combined with a RoBERTa text classifier for the textual modality. Subsequently, we build a multimodal classifier by combining text and audio representations. Evaluation on unseen test data demonstrates promising results, with Unweighted Average Recall scores reaching 82.46% in audio-only experiments, 85.97% in text-only experiments, and 87.16% using a multimodal approach.
Objective: Early diagnosis of cardiovascular diseases is a crucial task in medical practice. With the application of computer audition in the healthcare field, artificial intelligence (AI) has been applied to clinical non-invasive intelligent auscultation of heart sounds to provide rapid and effective pre-screening. However, AI models generally require large amounts of data which may cause privacy issues. Unfortunately, it is difficult to collect large amounts of healthcare data from a single centre. Methods: In this study, we propose federated learning (FL) optimisation strategies for the practical application in multi-centre institutional heart sound databases. The horizontal FL is mainly employed to tackle the privacy problem by aligning the feature spaces of FL participating institutions without information leakage. In addition, techniques based on deep learning have poor interpretability due to their “black-box” property, which limits the feasibility of AI in real medical data. To this end, vertical FL is utilised to address the issues of model interpretability and data scarcity. Conclusion: Experimental results demonstrate that, the proposed FL framework can achieve good performance for heart sound abnormality detection by taking the personal privacy protection into account. Moreover, using the federated feature space is beneficial to balance the interpretability of the vertical FL and the privacy of the data. Significance: This work realises the potential of FL from research to clinical practice, and is expected to have extensive application in the federated smart medical system.
The delayed access to specialized psychiatric assessments and care for patients at risk of suicidal tendencies in emergency departments creates a notable gap in timely intervention, hindering the provision of adequate mental health support during critical situations. To address this, we present a non-invasive, speech-based approach for automatic suicide risk assessment. For our study, we collected a novel speech recording dataset from 20 patients. We extract three sets of features, including wav2vec, interpretable speech and acoustic features, and deep learning-based spectral representations. We proceed by conducting a binary classification to assess suicide risk in a leave-one-subject-out fashion. Our most effective speech model achieves a balanced accuracy of 66.2%. Moreover, we show that integrating our speech model with a series of patients’ metadata, such as the history of suicide attempts or access to firearms, improves the overall result. The metadata integration yields a balanced accuracy of 94.4%, marking an absolute improvement of 28.2%, demonstrating the efficacy of our proposed approaches for automatic suicide risk assessment in emergency medicine.
Imbuing machines with the ability to talk has been a longtime pursuit of artificial intelligence (AI) research. From the very beginning, the community has not only aimed to synthesise high-fidelity speech that accurately conveys the semantic meaning of an utterance, but also to colour it with inflections that cover the same range of affective expressions that humans are capable of. After many years of research, it appears that we are on the cusp of achieving this when it comes to single, isolated utterances. This unveils an abundance of potential avenues to explore when it comes to combining these single utterances with the aim of synthesising more complex, longer-term behaviours. In the present chapter, we outline the methodological advances that brought us so far and sketch out the ongoing efforts to reach that coveted next level of artificial expressivity. We also discuss the societal implications coupled with rapidly advancing expressive speech synthesis (ESS) technology and highlight ways to mitigate those risks and ensure the alignment of ESS capabilities with ethical norms.
Recent studies have demonstrated the emerging capabilities of foundation models like ChatGPT in several fields, including affective computing. However, accessing these emerging capabilities is facilitated through prompt engineering. Despite the existence of some prompting techniques, the field is still rapidly evolving and many prompting ideas still require investigation. In this work, we introduce a method to evaluate and investigate the sensitivity of the performance of foundation models based on different prompts or generation parameters. We perform our evaluation on ChatGPT within the scope of affective computing on three major problems, namely sentiment analysis, toxicity detection, and sarcasm detection. First, we carry out a sensitivity analysis on pivotal parameters in auto-regressive text generation, specifically the temperature parameter T and the top-p parameter in Nucleus sampling, dictating how conservative or creative the model should be during generation. Furthermore, we explore the efficacy of several prompting ideas, where we explore how giving different incentives or structures affect the performance. Our evaluation takes into consideration performance measures on the affective computing tasks, and the effectiveness of the model to follow the stated instructions, hence generating easy-to-parse responses to be smoothly used in downstream applications.
We present the technologies and host components developed to power a speech-based dialogue manager with affective capabilities. The overall goal is that the system adapts its response to the sentiment and arousal level of the user inferred by analysing the linguistic and paralinguistic information embedded in his or her interaction. A linguistic-based, dedicated sentiment analysis component determines the body of the system response. A paralinguistic-based, dedicated arousal recognition component adjusts the energy level to convey in the affective system response. The sentiment analysis model is trained using the CMU-MOSEI dataset and implements a hierarchical contextual attention fusion network, which scores an Unweighted Average Recall (UAR) of 79.04% on the test set when tackling the task as a binary classification problem. The arousal recognition model is trained using the MSP-Podcast corpus. This model extracts the Mel-spectrogram representations of the speech signals, which are exploited with a Convolutional Neural Network (CNN) trained from scratch, and scores a UAR of 61.11% on the test set when tackling the task as a three-class classification problem. Furthermore, we highlight two sample dialogues implemented at the system back-end to detail how the sentiment and arousal inferences are coupled to determine the affective system response. These are also showcased in a proof of concept demonstrator. We publicly release the trained models to provide the research community with off-the-shelf sentiment analysis and arousal recognition tools.
The Internet of Things (IoT)-based passive acoustic monitoring (PAM) has shown great potential in large-scale remote bird monitoring. However, field recordings often contain overlapping signals, making precise bird information extraction challenging. To solve this challenge, first, the interchannel spatial feature is chosen as complementary information to the spectral feature to obtain additional spatial correlations between the sources. Then, an end-to-end model named BACPPNet is built based on Deeplabv3plus and enhanced with the polarized self-attention mechanism to estimate the spectral magnitude mask (SMM) for separating bird vocalizations. Finally, the separated bird vocalizations are recovered from SMMs and the spectrogram of mixed audio using the inverse short Fourier transform (ISTFT). We evaluate our proposed method utilizing the generated mixed data set. Experiments have shown that our method can separate bird vocalizations from mixed audio with root mean square error (RMSE), source-to-distortion ratio (SDR), source-to-interference ratio (SIR), source-to-artifact ratio (SAR), and short-time objective intelligibility (STOI) values of 2.82, 10.00 dB, 29.90 dB, 11.08 dB, and 0.66, respectively, which are better than existing methods. Furthermore, the average classification accuracy of the separated bird vocalizations drops the least. This indicates that our method outperforms other compared separation methods in bird sound separation and preserves the fidelity of the separated sound sources, which might help us better understand wild bird sound recordings.
©all images: LMU | TUM