Emotion Recognition in Human-Robot Interaction: Multimodal

Fusion, Deep Learning, and Ethical Considerations

Yuan Gao

College of Arts and Sciences, University of Washington, 4801 24th Ave NE APT 6130, Seattle, U.S.A.

Keywords: Emotion Recognition, Human-Robot Interaction, EEG-Based Emotion Detection, Multimodal Fusion,

Empathy in Robotics, Ethical AI.

Abstract: This paper explores recent advancements in emotion recognition techniques within Human-Robot Interaction

(HRI), focusing on the evolution of perception-based, electroencephalogram (EEG)-based, and multimodal

fusion approaches. Emotion recognition has become essential for enhancing the emotional intelligence of

robots, which can now better detect and respond to human emotions, particularly within service and healthcare

applications. Key contributions of this study include the integration of deep learning models, such as

Convolutional Neural Networks (CNNs) and Transformer-based architectures, which have shown significant

improvements in real-time emotion recognition accuracy. Additionally, the paper discusses the integration of

hardware innovations that optimize responsiveness, enabling robots to provide a more empathetic and

supportive experience. This research also examines the role of empathy within social robotics and its ethical

implications, covering data privacy, user consent, and the need for fair, unbiased artificial intelligence (AI).

By emphasizing the importance of regulatory frameworks, this study outlines the future of emotion-aware AI

in safe, ethical, and effective human-robot collaboration.

1 INTRODUCTION

As robots become an increasingly common presence

in our daily lives, the need for them to understand and

respond to human emotions has never been more

crucial. Imagine a healthcare robot comforting a

patient not just with physical assistance, but also with

emotional support, or a customer service robot that

can detect frustration and adjust its responses to calm

the situation. These scenarios, once considered

science fiction, are rapidly becoming reality due to

advancements in emotion recognition technology.

Homburg’s research in robotic psychology, over the

past two decades, has laid the foundation for these

developments by emphasizing the importance of

emotional interaction in enhancing human-robot

collaboration (Stock-Homburg, 2022). The question

is: What kind of an emotion recognition system could

we employ to boost Human-Robot Interaction (HRI)?

Accurately detecting and interpreting human

emotions is fundamental to enhancing robot

efficiency in interacting with humans. Emotion

recognition enables robots to adjust their behavior in

real-time, improving the perceived naturalness of

interactions and creating a more personalized

experience. This ability is especially crucial in HRI

scenarios that demand immediate emotional

responses, such as when assisting patients or

managing customer inquiries. However, according to

Marcos-Pablos and García-Peñalvo (2022), real-time

emotion detection remains a challenging task due to

the complexity of human emotions and the need to

integrate multimodal data such as facial expressions,

speech, and physiological signals (Marcos-Pablos &

García-Peñalvo, 2022). Techniques like multitask

learning have shown promise in addressing these

challenges by improving emotion detection accuracy

across various data modalities (Li, Kazemeini, Mehta

& Cambria, 2022).

The purpose of this review is to provide an

overview of the development and the latest

advancements in emotion recognition technologies,

particularly focusing on their application in HRI. This

paper will first examine traditional methods of

emotion detection, such as facial expression and

speech recognition. Next, it will explore cutting-edge

innovations like transformer-based models that offer

improved real-time performance and multimodal

systems that enhance detection accuracy across

diverse contexts (Heredia, Lopes-Silva, Cardinale,

Gao and Y.

Emotion Recognition in Human-Robot Interaction: Multimodal Fusion, Deep Learning, and Ethical Considerations.

DOI: 10.5220/0013526100004619

In Proceedings of the 2nd International Conference on Data Analysis and Machine Learning (DAML 2024), pages 459-465

ISBN: 978-989-758-754-2

459

Diaz-Amado, Dongo & Graterol, 2022). The review

concludes with a discussion of the ethical

considerations and future trends that will shape the

development of emotionally intelligent robots (Stark

& Hoey, 2021).

2 EMOTION DETECTION

TECHNIQUES IN HUMAN-

ROBOT INTERACTION

With the rapid development of human-robot

interaction in various fields, emotion recognition has

become a critical component in enhancing natural

interactions between humans and robots. Emotion

recognition not only enables robots to better

understand human emotions but also allows them to

adjust their behaviour based on the user’s emotional

state, which is particularly important in applications

such as healthcare, customer service, and

collaborative robots. Currently, emotion recognition

techniques in HRI primarily focus on three

categories: traditional perception-based detection

methods, electroencephalogram (EEG)-based

physiological emotion recognition, and multimodal

emotion detection techniques.

2.1 Traditional Emotion Recognition

Techniques

Traditional emotion recognition methods primarily

rely on observable external features such as facial

expressions, voice signals, and body language. Facial

expression recognition, which analyzes specific facial

muscle movements to infer emotional states, has been

widely used in emotion detection for HRI. Based on

Ekman’s facial expression theory, this method

identifies basic emotions like anger, joy, and sadness,

providing robots with intuitive emotional cues. For

example, Cavallo et al. (2018) have shown that facial

expressions can assist service robots in recognizing

customers' emotional responses, thereby improving

the quality of service (Cavallo et al., 2018).

However, facial expression recognition

techniques face limitations in complex or occluded

environments. Chuah and Yu (2021) shared that

facial expressions do not always accurately reflect an

individual’s true emotional state (Chuah & Yu, 2021).

Due to social norms or environmental pressures, users

may conceal their true emotional expressions.

Consequently, researchers have shifted toward

speech recognition techniques, analyzing tone,

speech rate, and volume to complement emotion

recognition accuracy. Emotional information

embedded in voice signals can significantly enhance

HRI, especially in scenarios where visual cues are

unavailable or unreliable (Rawal & Stock-Homburg,

2022).

While traditional facial expression and voice

recognition methods play a significant role in emotion

detection, their reliance on a single modality limits

their effectiveness in complex interaction scenarios,

motivating the exploration of more sophisticated

emotion recognition methods.

2.2 EEG-Based Emotion Recognition

To improve the accuracy and reliability of emotion

recognition, physiological signals have gained

attention in the field, particularly EEG-based emotion

recognition techniques. Emotion influences brain

activity at the neurological level, and EEG signals can

capture these changes in real time, making them a

powerful tool for emotion recognition (Cui et al.,

2020). EEG-based emotion recognition in HRI

benefits from its ability to detect subtle emotional

states, especially when facial expressions or voice

signals are not clearly exhibited.

In recent years, deep learning models for EEG-

based emotion recognition have shown significant

progress. For instance, Li, Meng, Wang and Hou

(2023) have proposed the Source-guided Multi-

Target Learning model, which processes the regional

differences in EEG signals to achieve high-accuracy

emotion recognition (Li, Meng, Wang & Hou, 2023).

Compared to traditional methods, this model captures

more complex emotional patterns and is suitable for

real-time applications in HRI where responsiveness is

critical. Furthermore, by combining EEG with

recurrent neural networks (RNN), Shen et al. (2020)

enhances the model’s ability to process temporal

sequences of data, improving both the speed and

accuracy of emotion detection (Shen et al., 2020).

Although EEG-based emotion recognition

surpasses traditional facial and voice recognition

methods in terms of accuracy, it still faces challenges.

Firstly, the wearable nature of EEG devices can affect

user comfort and natural interaction experiences.

Secondly, individual variability in brain signals

demands that emotion recognition models be highly

personalized and adaptive to achieve consistent

results.

DAML 2024 - International Conference on Data Analysis and Machine Learning

460

2.3 Multimodal Emotion Detection

Techniques

Multimodal emotion detection techniques integrate

various sensory inputs such as facial expressions,

voice, and EEG signals to enhance the accuracy and

robustness of emotion recognition (Xie, Sidulova &

Park, 2021). By combining information from

different modalities, multimodal approaches

overcome the limitations of single-modality methods.

For example, when visual signals are unstable or

speech is ambiguous, EEG signals can provide

supplementary information, ensuring continuous

emotion detection.

Recent studies have demonstrated that

multimodal approaches can significantly improve

emotion recognition accuracy. Xie, Sidulova and

Park’s (2021) study combined voice, text, and facial

expressions to develop a cross-modal emotion

detection model, greatly enhancing the accuracy of

emotion recognition in conversational settings (Park

et al., 2021). Last but not the least, the Multimodal

Fusion Network (MMFN), which integrates touch

gestures and facial expression data, introduced a

novel framework that substantially improved emotion

detection accuracy in complex interaction scenarios

(Li et al., 2023).

The Interactive Emotional Dyadic Motion

Capture (IEMOCAP) dataset is widely used for

evaluating multimodal emotion detection systems,

making it a benchmark for comparing methods like

speech, facial expression, text, and multimodal fusion

techniques. Table 1 illustrates the accuracy and

applicability of these emotion recognition methods

based on IEMOCAP results.

Table 1: Comparison of Emotion Detection Techniques on

IEMOCAP Dataset (Tripathi, Tripathi & Beigi, 2019)

Method Data

Type

Accurac

y (%)

Applicability

Facial

Emotion

Recognition

Motion

Capture

(Face,

Head,

Hand

)

48.99% Suitable for

clear facial

cues in well-

lit

environments.

Speech

Expression

Recognition

Audio

Signal

55.65% Effective for

remote

interactions or

limited visual

data.

Text

Emotion

Detection

Text

(Transcr

ipts)

64.78% Useful in

conversations

or textual

dialo

ues.

Multimodal

Fusion

Audio,

Text,

Motion

Signals

71.04% Optimal in

dynamic and

complex

interaction

environments.

Among different models, EEG stands out for its

high accuracy in detecting subtle emotional states that

are often missed by speech or facial cues, making it

invaluable for nuanced emotional analysis.

Multimodal approaches, which integrate signals from

various sources, consistently outperform single-

modality techniques. By leveraging speech, visual,

and physiological data, these systems offer improved

accuracy and adaptability, enabling more natural and

responsive human-robot interactions in dynamic

environments.

3 INNOVATIONS IN EMOTION

REGOGNITION

In recent years, the field of emotion recognition in

human-robot interaction has seen substantial growth

due to advancements in both algorithmic methods and

hardware developments. This chapter explores these

innovations, focusing on improvements in traditional

emotion recognition techniques, the application of

deep learning in multimodal systems, and the role of

hardware in enhancing real-time emotion detection.

3.1 Deep Learning in Multimodal

Emotion Recognition

The application of deep learning has greatly advanced

the field of emotion recognition, especially in systems

that combine multiple data modalities such as visual,

auditory, and physiological signals. Convolutional

neural networks (CNNs) and Long Short-Term

Memory (LSTM) networks have been pivotal in

extracting features from large, complex datasets,

improving the accuracy and adaptability of emotion

detection systems (Cui et al., 2020).

One significant innovation in this area is the use

of Transformer-based models, which outperform

traditional methods by capturing long-range

dependencies between different modalities. These

models have been applied to fuse data from speech,

text, and visual sources, thereby enhancing the

accuracy of emotion recognition in conversation-

based interactions (Park et al., 2021). According to

Park et al., the use of Transformers enabled real-time

integration of multiple data streams, allowing for

Emotion Recognition in Human-Robot Interaction: Multimodal Fusion, Deep Learning, and Ethical Considerations

461

more nuanced emotion detection in human-robot

interaction.

Multitask learning is another deep learning

technique that has been effectively utilized in emotion

recognition systems. It allows a single model to

simultaneously predict multiple emotional states or

personality traits, leading to better generalization

across tasks and more efficient use of data (Li et al.,

2022). These advancements are particularly relevant

in dynamic, real-time applications in HRI, where

systems must continuously adapt to varying

emotional cues from users.

3.2 Hardware Innovations for Emotion

Recognition

In addition to breakthroughs in algorithms,

advancements in hardware have played a crucial role

in improving the practicality of emotion recognition

systems. Wearable sensors, particularly those used

for capturing electroencephalography signals, have

become more lightweight and portable, making it

easier to integrate physiological data into emotion

recognition frameworks (Cui et al., 2020). These

sensors provide real-time data on brain activity,

which, when combined with deep learning

algorithms, allows for more accurate inference of

emotional states in naturalistic settings.

Figure 1 illustrates the process of EEG-based

emotion recognition, where brain activity is recorded

in response to emotional stimuli. The raw EEG data

is then preprocessed to remove artifacts, followed by

feature extraction of key brainwave frequencies.

These features are input into an emotion recognition

model, which classifies the emotional state of the user

(e.g., happy, neutral, sad). This integration of EEG

sensors and emotion recognition models enables real-

time, accurate detection of emotions, further

enhancing the effectiveness of human-robot

interaction in naturalistic environments.

Figure 1: Process of EEG-based

emotion recognition

Moreover, memristive circuits have emerged as a

powerful tool for enhancing hardware-based emotion

recognition. Memristive circuits simulate brain-like

learning processes, enabling more energy-efficient

processing of emotional data (Wang, Wang & Zeng,

2023). This innovation allows for the deployment of

emotion recognition systems on low-power devices,

such as robots, while maintaining high processing

speed. The development of brain-computer interfaces

(BCIs) has also improved the potential for seamless

human-robot interaction. By allowing robots to

access real-time emotional feedback from users via

direct neural signals, BCIs facilitate more intuitive

and empathetic interactions (Wang, Wang & Zeng,

2023).

4 INNOVATIONS IN EMOTION

REGOGNITION

4.1 Emotional Intelligence Through

Multimodal Integration

In modern human-robot interaction, robots are

required not only to recognize human emotions but

also to respond to them through integrated

multimodal feedback mechanisms. Emotionally

intelligent robots collect data from various sources

such as voice, facial expressions, and behaviors,

dynamically adjusting their responses to create a

more natural and emotionally sensitive interaction

experience. These systems enhance the user

experience by tailoring robot behavior based on real-

time emotional cues.

Hong et al. (2021) emphasizes the significance of

bidirectional emotional interaction between robots

and humans. Through the integration of multiple data

sources, robots can more accurately gauge users'

emotions and provide immediate, contextually

appropriate feedback (Hong et al., 2021). This real-

time emotional processing allows robots to adjust

their interaction strategies and align their behavior

with the emotional state of the user, improving both

engagement and user satisfaction. Additionally, Yan,

Iliyasu and Hirota (2021) highlights how robots can

utilize structured emotional representations to fine-

tune their responses, contributing to a more fluid

emotional exchange (Yan, Iliyasu & Hirota, 2021).

Artificial Emotional Intelligence (AEI) ought to seek

to simulate and amplify natural emotions—

particularly human emotions—to equip robots with

the ability to recognize and express emotions during

HRI.

DAML 2024 - International Conference on Data Analysis and Machine Learning

462

4.2 The Role of Empathy and Emotional

Feedback in Social Robots

As social robots evolve, empathy has become a

crucial component in improving the quality of

human-robot interactions. Robots that can understand

and respond empathetically to human emotions are

more likely to be accepted and integrated into fields

such as healthcare, education, and customer service.

Park and Whang (2022) present a framework for

developing empathetic systems in social robots. This

framework focuses on how robots can enhance

emotional connections with users through effective

emotion recognition and feedback (Park & Whang,

2022). Additionally, Chiang, Trimi and Lo (2022)

discusses how emotional feedback in service robots

improves the overall user experience, demonstrating

that empathetic robots can significantly enhance the

perceived quality of interactions (Chiang, Trimi &

Lo, 2022). By responding to users' emotions, these

robots can foster deeper emotional bonds and

improve satisfaction across a variety of applications.

5 APPLICATIONS AND FUTURE

TRENDS

5.1 Service Robots

Emotional Artificial Intelligence (AI) plays a crucial

role in the advancement of service robots, particularly

in customer service and personalized interaction.

Emotional intelligence in robots allows them to

recognize and adapt to users’ emotional states,

thereby enhancing user satisfaction and making

interactions more meaningful. This capability is

essential in environments such as retail, hospitality,

and healthcare, where personalized service is critical

to improving user experiences.

Chuah and Yu (2021) emphasize that emotion-

aware robots significantly improve the quality of

service by responding to customers’ emotional needs

and tailoring interactions accordingly. A robot

capable of detecting customer frustration can adjust

its behavior to offer more supportive responses.

Similarly, by employing edge clouds on smart

clothing, Yang et al. (2020) highlights how emotional

intelligence in robots contributes to a deeper

engagement, enabling them to build rapport with

users by offering more human-like interaction (Yang

et al., 2020). These advancements make service

robots more effective at handling complex customer

needs, fostering trust, and ensuring long-term user

satisfaction.

5.2 Collaborative Robots and Safety

Emotion recognition can also enhance the safety and

efficiency of collaborative robots, especially in

industrial and workplace settings. By understanding

and responding to human emotional cues,

collaborative robots can adjust their actions to prevent

accidents or discomfort, contributing to a safer

working environment. Emotional AI allows these

robots to anticipate and react to potentially dangerous

situations, such as stress or frustration in human

coworkers, which may otherwise lead to mistakes or

unsafe behavior.

Toichoa, Mohammed and Martinez (2021)

demonstrate how emotion recognition helps

collaborative robots better understand human intent

and adjust their operational behavior to enhance

safety and collaboration (Toichoa, Mohammed &

Martinez, 2021). Additionally, Zacharaki, Kostavelis,

Gasteratos and Dokas (2020) explore how integrating

emotional awareness into robots’ control systems can

help set safety parameters that minimize risk while

maximizing operational efficiency (Zacharaki,

Kostavelis, Gasteratos & Dokas, 2020). By

incorporating emotional cues, collaborative robots

become more aware of their human coworkers’

needs, making teamwork more fluid and secure.

5.3 Ethical Considerations

As emotional AI becomes more prevalent, ethical

concerns surrounding privacy, emotional

manipulation, and data security have become critical

issues. Emotionally intelligent robots process

sensitive emotional data, which raises concerns about

how this information is collected, stored, and used.

The potential for emotional manipulation—where

robots could influence users' emotions in unethical

ways—also poses a significant ethical challenge,

particularly in vulnerable populations such as

children, the elderly, or individuals with cognitive

impairments.

Stark and Hoey (2021) addresses these concerns

by highlighting the need for strict regulatory

frameworks to ensure emotional AI systems respect

user privacy and autonomy. There are also concerns

about emotional manipulation, especially when

robots are used with vulnerable populations such as

children, the elderly, or individuals with cognitive

impairments. The ethical design of emotion-aware

systems must prioritize the well-being of users,

ensuring that emotional feedback is used responsibly

Emotion Recognition in Human-Robot Interaction: Multimodal Fusion, Deep Learning, and Ethical Considerations

463

and without crossing personal boundaries. Ensuring

transparency and user consent will be vital as

emotional AI becomes more integrated into daily life.

6 CONCLUSIONS

Emotion recognition has become a critical aspect of

human-robot interaction, with profound implications

for how robots understand and respond to human

needs. This review has examined the key

advancements in emotion detection technologies,

highlighting the significance of both traditional

methods and more innovative approaches such as

multimodal integration and deep learning techniques.

The use of multimodal data, including facial

expressions, speech, and physiological signals, has

substantially improved the accuracy and reliability of

emotion recognition systems, enabling robots to

adapt in real time and enhance the naturalness of

human-robot interactions.

Despite the progress made, there are still

significant challenges that need to be addressed. Real-

time processing of complex, multimodal emotional

data remains a technical hurdle, particularly in

environments where quick adaptation is required.

Additionally, understanding and responding to

emotions in a way that is culturally sensitive and

contextually appropriate poses ongoing difficulties.

Addressing these challenges will require not only

advances in algorithms and hardware but also a

deeper exploration of the psychological and social

dimensions of emotion in human-robot interaction.

Looking forward, the future of emotional

intelligent robots lies in further refining these

technologies to create more seamless, empathetic

interactions. This includes the development of more

adaptive and personalized emotion recognition

models that can operate across diverse environments

and user groups. Moreover, ethical considerations

must be an integral part of future research, ensuring

that the deployment of emotion-aware robots

promotes positive human experiences without

infringing on privacy or autonomy.

REFERENCES

Cavallo, F., Semeraro, F., Fiorini, L., Magyar, G., Sinčák,

P., & Dario, P. (2018). Emotion modelling for Social

Robotics Applications: A Review. Journal of Bionic

Engineering, 15(2), 185–203.

https://doi.org/10.1007/s42235-018-0015-y

Chiang, A.-H., Trimi, S., & Lo, Y.-J. (2022). Emotion and

service quality of anthropomorphic robots.

Technological Forecasting and Social Change, 177,

121550.

https://doi.org/10.1016/j.techfore.2022.121550

Chuah, S. H.-W., & Yu, J. (2021). The future of service:

The power of emotion in human-robot interaction.

Journal of Retailing and Consumer Services, 61,

102551.

https://doi.org/10.1016/j.jretconser.2021.102551

Cui, H., Liu, A., Zhang, X., Chen, X., Wang, K., & Chen,

X. (2020). EEG-based emotion recognition using an

end-to-end regional-asymmetric convolutional neural

network. Knowledge-Based Systems, 205, 106243.

https://doi.org/10.1016/j.knosys.2020.106243

Heredia, J., Lopes-Silva, E., Cardinale, Y., Diaz-Amado, J.,

Dongo, I., Graterol, W., & Aguilera, A. (2022).

Adaptive Multimodal Emotion Detection Architecture

for Social Robots. IEEE Access, 10, 20727–20744.

https://doi.org/10.1109/access.2022.3149214

Hong, A., Lunscher, N., Hu, T., Tsuboi, Y., Zhang, X.,

Franco dos Reis Alves, S., Nejat, G., & Benhabib, B.

(2021). A multimodal emotional human–robot

interaction architecture for social robots engaged in

bidirectional communication. IEEE Transactions on

Cybernetics, 51(12), 5954–5968.

https://doi.org/10.1109/tcyb.2020.2974688

Li, Y., Kazemeini, A., Mehta, Y., & Cambria, E. (2022).

Multitask learning for emotion and personality traits

detection. Neurocomputing, 493, 340–350.

https://doi.org/10.1016/j.neucom.2022.04.049

Li, Y.-K., Meng, Q.-H., Wang, Y.-X., & Hou, H.-R. (2023).

MMFN: Emotion recognition by fusing touch gesture

and facial expression information. Expert Systems with

Applications, 228, 120469.

https://doi.org/10.1016/j.eswa.2023.120469

Marcos-Pablos, S., & García-Peñalvo, F. J. (2021).

Emotional intelligence in robotics: A scoping review.

Advances in Intelligent Systems and Computing, 66–75.

https://doi.org/10.1007/978-3-030-87687-6_7

Park, S., & Whang, M. (2022). Empathy in human–robot

interaction: Designing for social robots. International

Journal of Environmental Research and Public Health,

19(3), 1889. https://doi.org/10.3390/ijerph19031889

Rawal, N., & Stock-Homburg, R. M. (2022). Facial

emotion expressions in human–robot interaction: A

survey. International Journal of Social Robotics, 14(7),

1583–1604. https://doi.org/10.1007/s12369-022-

00867-0

Shen, F., Dai, G., Lin, G., Zhang, J., Kong, W., & Zeng, H.

(2020). EEG-based emotion recognition using 4D

convolutional recurrent neural network. Cognitive

Neurodynamics, 14(6), 815–828.

https://doi.org/10.1007/s11571-020-09634-1

Stark, L., & Hoey, J. (2021). The ethics of Emotion in

Artificial Intelligence Systems. Proceedings of the

2021 ACM Conference on Fairness, Accountability,

and Transparency.

https://doi.org/10.1145/3442188.3445939

DAML 2024 - International Conference on Data Analysis and Machine Learning

464

Stock-Homburg, R. (2021). Survey of emotions in human–

robot interactions: Perspectives from robotic

psychology on 20 years of research. International

Journal of Social Robotics, 14(2), 389–411.

https://doi.org/10.1007/s12369-021-00778-6

Toichoa Eyam, A., Mohammed, W. M., & Martinez Lastra,

J. L. (2021). Emotion-driven analysis and control of

human-robot interactions in collaborative applications.

Sensors, 21(14), 4626.

https://doi.org/10.3390/s21144626

Tripathi, S., Tripathi, S., & Beigi, H. (2019a, November 6).

Multi-modal emotion recognition on IEMOCAP dataset

using Deep Learning. arXiv.org.

https://arxiv.org/abs/1804.05788

Wang, Z., Wang, X., & Zeng, Z. (2023). Memristive circuit

design of brain-like emotional learning and generation.

IEEE Transactions on Cybernetics, 53(1), 222–235.

https://doi.org/10.1109/tcyb.2021.3090811

Xie, B., Sidulova, M., & Park, C. H. (2021). Robust

multimodal emotion recognition from conversation

with Transformer-based crossmodality fusion. Sensors,

21(14), 4913. https://doi.org/10.3390/s21144913

Yan, F., Iliyasu, A. M., & Hirota, K. (2021). Emotion space

modelling for social robots. Engineering Applications

of Artificial Intelligence, 100, 104178.

https://doi.org/10.1016/j.engappai.2021.104178

Yang, J., Wang, R., Guan, X., Hassan, M. M., Almogren,

A., & Alsanad, A. (2020). Ai-enabled emotion-aware

robot: The fusion of Smart Clothing, edge clouds and

robotics. Future Generation Computer Systems, 102,

701–709. https://doi.org/10.1016/j.future.2019.09.029

Zacharaki, A., Kostavelis, I., Gasteratos, A., & Dokas, I.

(2020). Safety Bounds in human robot interaction: A

survey. Safety Science, 127, 104667.

https://doi.org/10.1016/j.ssci.2020.104667

Emotion Recognition in Human-Robot Interaction: Multimodal Fusion, Deep Learning, and Ethical Considerations

465