Engin Erzin

Ali Safaya, Engin Erzin

While the Turkish language is listed among low-resource languages, literature on Turkish automatic speech recognition (ASR) is relatively old. In this report, we present our findings on Turkish ASR with speech representation learning using HUBERT. We investigate pre-training HUBERT for Turkish with large-scale data curated from online resources. We pre-train our model using 6,500 hours of speech data from YouTube. The results show that the models are not ready for commercial use since they are not robust against disturbances that typically occur in real-world settings such as variations in accents, slang, background noise and interference. We analyze typical errors and the limitations of the models for use in commercial settings.

Ibrahim Shoer, Berkay Kopru, Engin Erzin

Video summarization attracts attention for efficient video representation, retrieval, and browsing to ease volume and traffic surge problems. Although video summarization mostly uses the visual channel for compaction, the benefits of audio-visual modeling appeared in recent literature. The information coming from the audio channel can be a result of audio-visual correlation in the video content. In this study, we propose a new audio-visual video summarization framework integrating four ways of audio-visual information fusion with GRU-based and attention-based networks. Furthermore, we investigate a new explainability methodology using audio-visual canonical correlation analysis (CCA) to better understand and explain the role of audio in the video summarization task. Experimental evaluations on the TVSum dataset attain F1 score and Kendall-tau score improvements for the audio-visual video summarization. Furthermore, splitting video content on TVSum and COGNIMUSE datasets based on audio-visual CCA as positively and negatively correlated videos yields a strong performance improvement over the positively correlated videos for audio-only and audio-visual video summarization.

Ibrahim Shoer, Engin Erzin

In affective computing, datasets often contain multiple annotations from different annotators, which may lack full agreement. Typically, these annotations are merged into a single gold standard label, potentially losing valuable inter-rater variability. We propose a multi-annotator training approach for continuous emotion recognition (CER) that seeks a consensus across all annotators rather than relying on a single reference label. Our method employs a consensus network to aggregate annotations into a unified representation, guiding the main arousal-valence predictor to better reflect collective inputs. Tested on the RECOLA and COGNIMUSE datasets, our approach outperforms traditional methods that unify annotations into a single label. This underscores the benefits of fully leveraging multi-annotator data in emotion recognition and highlights its applicability across various fields where annotations are abundant yet inconsistent.

Berkay Kopru, Engin Erzin

As speech-interfaces are getting richer and widespread, speech emotion recognition promises more attractive applications. In the continuous emotion recognition (CER) problem, tracking changes across affective states is an important and desired capability. Although CER studies widely use correlation metrics in evaluations, these metrics do not always capture all the high-intensity changes in the affective domain. In this paper, we define a novel affective burst detection problem to accurately capture high-intensity changes of the affective attributes. For this problem, we formulate a two-class classification approach to isolate affective burst regions over the affective state contour. The proposed classifier is a kernel-fusion dilated convolutional neural network (KFDCNN) architecture driven by speech spectral features to segment the affective attribute contour into idle and burst sections. Experimental evaluations are performed on the RECOLA and CreativeIT datasets. The proposed KFDCNN is observed to outperform baseline feedforward neural networks on both datasets.

Berkay Köprü, Engin Erzin

Increasing volume of user-generated human-centric video content and their applications, such as video retrieval and browsing, require compact representations that are addressed by the video summarization literature. Current supervised studies formulate video summarization as a sequence-to-sequence learning problem and the existing solutions often neglect the surge of human-centric view, which inherently contains affective content. In this study, we investigate the affective-information enriched supervised video summarization task for human-centric videos. First, we train a visual input-driven state-of-the-art continuous emotion recognition model (CER-NET) on the RECOLA dataset to estimate emotional attributes. Then, we integrate the estimated emotional attributes and the high-level representations from the CER-NET with the visual information to define the proposed affective video summarization architectures (AVSUM). In addition, we investigate the use of attention to improve the AVSUM architectures and propose two new architectures based on temporal attention (TA-AVSUM) and spatial attention (SA-AVSUM). We conduct video summarization experiments on the TvSum database. The proposed AVSUM-GRU architecture with an early fusion of high level GRU embeddings and the temporal attention based TA-AVSUM architecture attain competitive video summarization performances by bringing strong performance improvements for the human-centric videos compared to the state-of-the-art in terms of F-score and self-defined face recall metrics.

Zana Bucinca, Yucel Yemez, Engin Erzin et al.

Due to its expressivity, natural language is paramount for explicit and implicit affective state communication among humans. The same linguistic inquiry (e.g., How are you?) might induce responses with different affects depending on the affective state of the conversational partner(s) and the context of the conversation. Yet, most dialog systems do not consider affect as constitutive aspect of response generation. In this paper, we introduce AffectON, an approach for generating affective responses during inference. For generating language in a targeted affect, our approach leverages a probabilistic language model and an affective space. AffectON is language model agnostic, since it can work with probabilities generated by any language model (e.g., sequence-to-sequence models, neural language models, n-grams). Hence, it can be employed for both affective dialog and affective language generation. We experimented with affective dialog generation and evaluated the generated text objectively and subjectively. For the subjective part of the evaluation, we designed a custom user interface for rating and provided recommendations for the design of such interfaces. The results, both subjective and objective demonstrate that our approach is successful in pulling the generated language toward the targeted affect, with little sacrifice in syntactic coherence.

Berkay Köprü, Engin Erzin

In this study, we focus on continuous emotion recognition using body motion and speech signals to estimate Activation, Valence, and Dominance (AVD) attributes. Semi-End-To-End network architecture is proposed where both extracted features and raw signals are fed, and this network is trained using multi-task learning (MTL) rather than the state-of-the-art single task learning (STL). Furthermore, correlation losses, Concordance Correlation Coefficient (CCC) and Pearson Correlation Coefficient (PCC), are used as an optimization objective during the training. Experiments are conducted on CreativeIT and RECOLA database, and evaluations are performed using the CCC metric. To highlight the effect of MTL, correlation losses and multi-modality, we respectively compare the performance of MTL against STL, CCC loss against root mean square error (MSE) loss and, PCC loss, multi-modality against single modality. We observe significant performance improvements with MTL training over STL, especially for estimation of the valence. Furthermore, the CCC loss achieves more than 7% CCC improvements on CreativeIT, and 13% improvements on RECOLA against MSE loss.

Rizwan Sadiq, Sasan AsadiAbadi, Engin Erzin

In human-to-computer interaction, facial animation in synchrony with affective speech can deliver more naturalistic conversational agents. In this paper, we present a two-stage deep learning approach for affective speech driven facial shape animation. In the first stage, we classify affective speech into seven emotion categories. In the second stage, we train separate deep estimators within each emotion category to synthesize facial shape from the affective speech. Objective and subjective evaluations are performed over the SAVEE dataset. The proposed emotion dependent facial shape model performs better in terms of the Mean Squared Error (MSE) loss and in generating the landmark animations, as compared to training a universal model regardless of the emotion.

Nusrah Hussain, Engin Erzin, T. Metin Sezgin et al.

The ability to generate appropriate verbal and non-verbal backchannels by an agent during human-robot interaction greatly enhances the interaction experience. Backchannels are particularly important in applications like tutoring and counseling, which require constant attention and engagement of the user. We present here a method for training a robot for backchannel generation during a human-robot interaction within the reinforcement learning (RL) framework, with the goal of maintaining high engagement level. Since online learning by interaction with a human is highly time-consuming and impractical, we take advantage of the recorded human-to-human dataset and approach our problem as a batch reinforcement learning problem. The dataset is utilized as a batch data acquired by some behavior policy. We perform experiments with laughs as a backchannel and train an agent with value-based techniques. In particular, we demonstrate the effectiveness of recurrent layers in the approximate value function for this problem, that boosts the performance in partially observable environments. With off-policy policy evaluation, it is shown that the RL agents are expected to produce more engagement than an agent trained from imitation learning.

Nusrah Hussain, Engin Erzin, T. Metin Sezgin et al.

We present a novel method for training a social robot to generate backchannels during human-robot interaction. We address the problem within an off-policy reinforcement learning framework, and show how a robot may learn to produce non-verbal backchannels like laughs, when trained to maximize the engagement and attention of the user. A major contribution of this work is the formulation of the problem as a Markov decision process (MDP) with states defined by the speech activity of the user and rewards generated by quantified engagement levels. The problem that we address falls into the class of applications where unlimited interaction with the environment is not possible (our environment being a human) because it may be time-consuming, costly, impracticable or even dangerous in case a bad policy is executed. Therefore, we introduce deep Q-network (DQN) in a batch reinforcement learning framework, where an optimal policy is learned from a batch data collected using a more controlled policy. We suggest the use of human-to-human dyadic interaction datasets as a batch of trajectories to train an agent for engaging interactions. Our experiments demonstrate the potential of our method to train a robot for engaging behaviors in an offline manner.