Filters








8,953 Hits in 4.8 sec

Improving Emotion Classification through Variational Inference of Latent Variables

Srinivas Parthasarathy, Viktor Rozgic, Ming Sun, Chao Wang
2019 ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)  
We propose an Adversarial Autoencoder (AAE) to perform variational inference over the latent variables and reconstruct the input feature representations.  ...  Experiments on the IEMOCAP dataset demonstrate that the auxiliary learning tasks improve emotion classification accuracy compared to a baseline supervised classifier.  ...  We pose the recognition of speech emotion as a latent variable inference problem and solve it using a variational inference procedure.  ... 
doi:10.1109/icassp.2019.8682823 dblp:conf/icassp/ParthasarathyRS19 fatcat:3badcjluyvgpnfsaiv7rjw6qlu

Animating Face using Disentangled Audio Representations

Gaurav Mittal, Baoyuan Wang
2020 2020 IEEE Winter Conference on Applications of Computer Vision (WACV)  
noise and emotional variations.  ...  To the best of our knowledge, this is the first work which improves the performance of talking head generation through a disentangled audio representation perspective, which is important for many real-world  ...  For instance, we enforce the content latent variable z i,n c for audio segment x i,n to correctly infer its viseme annotation v i,n through classification loss given by, log p(v i,n |z i,n c ) = log p(  ... 
doi:10.1109/wacv45572.2020.9093527 dblp:conf/wacv/MittalW20 fatcat:cdiijakmnnbctf4c4msnanrddm

Variational Autoencoders for Learning Latent Representations of Speech Emotion: A Preliminary Study [article]

Siddique Latif, Rajib Rana, Junaid Qadir, Julien Epps
2020 arXiv   pre-print
To the best of our knowledge, we are the first to propose VAEs for speech emotion classification.  ...  Inspired by this, we propose VAEs for deriving the latent representation of speech signals and use this representation to classify emotions.  ...  The idea of VAE is to infer P (z) using P (z|X), where P (z|X) is determined using Variational Inference (VI).  ... 
arXiv:1712.08708v3 fatcat:goyfbnmezzhttkpto746u5gnjm

Animating Face using Disentangled Audio Representations [article]

Gaurav Mittal, Baoyuan Wang
2019 arXiv   pre-print
noise and emotional variations.  ...  To our best knowledge, this is the first work which improves the performance of talking head generation from disentangled audio representation perspective, which is important for many real-world applications  ...  For instance, we enforce the content latent variable z i,n c for audio segment x i,n to correctly infer its viseme annotation v i,n through classification loss given by, log p(v i,n |z i,n c ) = log p(  ... 
arXiv:1910.00726v1 fatcat:ofmxiseyhnaxneam4mnz55atvu

Variational Autoencoders for Learning Latent Representations of Speech Emotion: A Preliminary Study

Siddique Latif, Rajib Rana, Junaid Qadir, Julien Epps
2018 Interspeech 2018  
To the best of our knowledge, we are the first to propose VAEs for speech emotion classification.  ...  Inspired by this, we propose VAEs for deriving the latent representation of speech signals and use this representation to classify emotions.  ...  The idea of VAE is to infer P (z) using P (z|X), where P (z|X) is determined using Variational Inference (VI).  ... 
doi:10.21437/interspeech.2018-1568 dblp:conf/interspeech/LatifRQE18 fatcat:f5obguwfsfgoxfidi3uqp5yt24

Deep Encoder-Decoder Models for Unsupervised Learning of Controllable Speech Synthesis [article]

Gustav Eje Henter, Jaime Lorenzo-Trueba, Xin Wang, Junichi Yamagishi
2018 arXiv   pre-print
For example, we show that popular unsupervised training heuristics can be interpreted as variational inference in certain autoencoder models.  ...  We illustrate the utility of the various approaches with an application to acoustic modelling for emotional speech synthesis, where the unsupervised methods for learning expression control (without access  ...  III-A outlines controllable speech synthesis through latent variables, while remaining sections describe the fundamental theory of variational inference (Sec.  ... 
arXiv:1807.11470v3 fatcat:2hqdissmirbrvoh4iu353fhzmi

Latent Space Energy-Based Model of Symbol-Vector Coupling for Text Generation and Classification [article]

Bo Pang, Ying Nian Wu
2021 arXiv   pre-print
We propose a latent space energy-based prior model for text generation and classification.  ...  The energy term of the prior model couples a continuous latent vector and a symbolic one-hot vector, so that discrete category can be inferred from the observed example based on the continuous latent vector  ...  latent variables in our models through an EBM is more expressive.  ... 
arXiv:2108.11556v1 fatcat:bg7jbweeh5ao3ai4xfveoxdm5i

ReCAB-VAE: Gumbel-Softmax Variational Inference Based on Analytic Divergence [article]

Sangshin Oh, Seyun Um, Hong-Goo Kang
2022 arXiv   pre-print
Along with this new metric, we propose a relaxed categorical analytic bound variational autoencoder (ReCAB-VAE) that successfully models both continuous and relaxed discrete latent representations.  ...  The Gumbel-softmax distribution, or Concrete distribution, is often used to relax the discrete characteristics of a categorical distribution and enable back-propagation through differentiable reparameterization  ...  The relaxed categorical distribution is frequently applied for variational inference [7] [8] [9] [10] [11] [12] , in which a distribution of latent representation estimated from observed data (a variational  ... 
arXiv:2205.04104v1 fatcat:7unqlr724fazpdhjgtkb622giu

Emotional Dialogue Generation Based on Conditional Variational Autoencoder and Dual Emotion Framework

Zhenrong Deng, Hongquan Lin, Wenming Huang, Rushi Lan, Xiaonan Luo, Yaguang Lin
2020 Wireless Communications and Mobile Computing  
In our model, latent variables of the conditional variational autoencoder are adopted to promote the diversity of conversation.  ...  A large number of experiments show that our model not only generates rich and diverse responses but also is emotionally coherent and controllable.  ...  AB20238013, ZY20198016, 2019GXNSFFA245014), and Guangxi Key Laboratory of Image and Graphic Intelligent Processing Project (No. GIIP2003).  ... 
doi:10.1155/2020/8881616 fatcat:ab7neiuoiravlfi2jztm576puu

Latent Factor Decoding of Multi-Channel EEG for Emotion Recognition Through Autoencoder-Like Neural Networks

Xiang Li, Zhigang Zhao, Dawei Song, Yazhou Zhang, Jingshan Pan, Lu Wu, Jidong Huo, Chunyang Niu, Di Wang
2020 Frontiers in Neuroscience  
Hence, the states of the latent variables that relate to emotional processing must contribute to building robust recognition models.  ...  Through a sequence modeling method, we examine the emotion recognition performance based on the learnt latent factors.  ...  Peng Zhang for taking care of our research work. Finally, we would also like to thank the editors, reviewers, editorial staffs who take part in the publication process of this paper.  ... 
doi:10.3389/fnins.2020.00087 pmid:32194367 pmcid:PMC7061897 fatcat:u2oe4ic6svaffgn7t5kzlcxcg4

Latent Emotion Memory for Multi-Label Emotion Classification

Hao Fei, Yue Zhang, Yafeng Ren, Donghong Ji
2020 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
In this paper, we propose a Latent Emotion Memory network (LEM) for multi-label emotion classification.  ...  The proposed model can learn the latent emotion distribution without external knowledge, and can effectively leverage it into the classification network.  ...  Yu et al. (2018) proposed a transfer learning architecture to improve the performance of multi-label emotion classification.  ... 
doi:10.1609/aaai.v34i05.6271 fatcat:rdm66lb47bcunjzxldxl2pt5p4

Context-LGM: Leveraging Object-Context Relation for Context-Aware Object Recognition [article]

Mingzhou Liu, Xinwei Sun, Fandong Zhang, Yizhou Yu, Yizhou Wang
2021 arXiv   pre-print
Specifically, we firstly introduce a latent generative model with a pair of correlated latent variables to respectively model the object and context, and embed their correlation via the generative process  ...  Then, to infer contextual features, we reformulate the objective function of Variational Auto-Encoder (VAE), where contextual features are learned as a posterior distribution conditioned on the object.  ...  " # Inferring (b) Latent Inference Process where q φ (z|x) is the variational distribution.  ... 
arXiv:2110.04042v1 fatcat:zlxfibelszfevctbwscgajzu6q

Semi-supervised Bayesian Deep Multi-modal Emotion Recognition [article]

Changde Du, Changying Du, Jinpeng Li, Wei-long Zheng, Bao-liang Lu, Huiguang He
2017 arXiv   pre-print
By imposing a mixture of Gaussians assumption on the posterior approximation of the latent variables, our model can learn the shared deep representation from multiple modalities.  ...  In emotion recognition, it is difficult to recognize human's emotional states using just a single modality. Besides, the annotation of physiological emotional data is particularly expensive.  ...  Overall, the model can be trained with reparameterization trick for backpropagation through the mixed Gaussian latent variables.  ... 
arXiv:1704.07548v1 fatcat:o334v3e6kvceplx5ouimcg4xzm

AutoBayes: Automated Bayesian Graph Exploration for Nuisance-Robust Inference

Andac Demir, Toshiaki Koike-Akino, Ye Wang, Deniz Erdogmus
2021 IEEE Access  
Auto-Bayes also enables learning disentangled representations, where the latent variable is split into multiple pieces to impose various relationships with the nuisance variation and task labels.  ...  We benchmark the framework on several public datasets, and provide analysis of its capability for subject-transfer learning with/without variational modeling and adversarial training.  ...  We further improved the performance approaching the state-of-the-art accuracy by exploiting multiple inference models explored in Auto-Bayes through the use of ensemble stacking.  ... 
doi:10.1109/access.2021.3064530 fatcat:hfwenaojunegfbytlcvc73z2e4

BIVA: A Very Deep Hierarchy of Latent Variables for Generative Modeling [article]

Lars Maaløe, Marco Fraccaro, Valentin Liévin, Ole Winther
2019 arXiv   pre-print
With the introduction of the variational autoencoder (VAE), probabilistic latent variable models have received renewed attention as powerful generative models.  ...  We attribute this to the hierarchy of latent variables which is able to extract high-level semantic features.  ...  The TD variables depend on the data and the BU variables lower in the hierarchy through the BU inference path, but also on all variables above in the hierarchy through the TD inference path, see Figure  ... 
arXiv:1902.02102v3 fatcat:trarar7u5fdepeqo7qb7vz6pjy
« Previous Showing results 1 — 15 out of 8,953 results