=Paper=
{{Paper
|id=Vol-1803/paper7
|storemode=property
|title=Emotion-Recognition from Speech-based
Interaction in AAL Environment
|pdfUrl=https://ceur-ws.org/Vol-1803/paper7.pdf
|volume=Vol-1803
|authors=Berardina De Carolis,Stefano Ferilli,Giuseppe Palestra,Domenico Redavid
|dblpUrl=https://dblp.org/rec/conf/aiia/CarolisFPR16
}}
==Emotion-Recognition from Speech-based
Interaction in AAL Environment==
https://ceur-ws.org/Vol-1803/paper7.pdf
Emotion-Recognition from Speech-based
Interaction in AAL Environment
B. De Carolis, S. Ferilli, G. Palestra, D. Redavid
Dipartimento di Informatica, Università di Bari
70126 Bari, Italy
@uniba.it
Abstract. In Ambient Assisted Living environments assistance and care are
delegated to the intelligence embedded in the environment that, in our opinion,
should provide not only a task-oriented support but also an interface able to es-
tablish a social empathic relation with the user. To this aim social assistive ro-
bots are being employed as a mediator interface and, in order to achieve a rela-
tion with the user, they should be endowed with the capability of recognizing
the user affective state. Since a natural way to interact with a robot is speech,
spoken user’s input can be used to give to the robot the capability of recogniz-
ing the emotions and attitude of the user, thus providing more detail infor-
mation about the user state. This paper focuses on this topic and proposes an
approach based on the dimensional model of emotions in which the valence and
arousal of user’s spoken input are recognized. The experimental analysis shows
the performance in terms of accuracy of the proposed approach on an Italian da-
taset. In order to show its application in the context of Ambient Assisted Liv-
ing, an example is provided.
1 Introduction
A Smart Environment should support people in their daily activities by assisting and
facilitating users when interacting with environment services in a natural and easy
way. The required assistance may be provided to the user through different devices.
The choice of an assistive robot agent as an interface is supported by several consid-
erations. First of all, the robot has a physical presence and it may participate in the
user’s daily life. Assistive robots can move around and perform actions, follow and
observe the user in the environment, which is fundamental when designing supportive
technologies for elderly people [1]. In addition to typical service-oriented features,
assistive robots can be equipped with social and conversational capabilities, thus im-
proving the naturalness and effectiveness of the interaction between users and smart
environment services.
Speech is a natural way for humans to interact with robots [2]. Moreover, speech
based interaction is seen as an effective interface for smart environments because it is
natural, hands-free and it enables different types of users with different capabilities,
and disabilities, to interact with systems. Since elderly people are an important user
group for smart environments, spoken interaction is of particular benefit for them
since it is natural and does not require particular skills. In addition, spoken user’s in-
put can be used not only to issue commands, but also to give to the robot the capabil-
ity of recognizing the emotions and attitude of the user and this is very important for
establishing a social relation and to personalize service execution. Indeed, providing
personalized services requires taking into account several factors, which are related to
the nature of the service, to user’s preferences and to context-related features such as
user’s emotional state.
In this paper we focus on the latter issue and we will present an acoustic analyzer
for the recognition of the emotion. This module is able to extract the prosodic features
of user’s spoken input and, starting from them is able to recognize the two dimension
of emotions: valence and arousal [25]. Then, the module has been used by a social
assistive robot embodied in NAO. The robot acts as Interactor Agent in a smart home
environment implemented as a Multi Agent System (MAS) [3].
The experimental analysis shows the performance in terms of accuracy of the pro-
posed approach on an Italian dataset. The obtained results also show which combina-
tion of features assures a satisfying recognition rate allowing a better understanding of
the user’s affective state.
The paper is structured as follows. In Section 2 the motivations and technical
background for this work are presented. Section 3 briefly describes the MAS architec-
ture implementing the smart home environment. Section 4 describes how VOCE has
been developed and Section 5 shows an example on how it can be applied in the con-
text of AAL. Conclusions and future work directions are illustrated in Section 6.
2 Background and Motivations
Interaction with services provided by a smart environment may be provided to the us-
er in a seamless way (i.e. by combining smart home technologies based on sensors
and effectors embedded in the appliances of the environment), or using an embodied
companion as an interface, or combining both approaches. In all cases, research em-
phasizes the need of natural and user-friendly interfaces for accessing the services
provided by the environment. Moreover, research on social and affective computing
suggests that such an assistive environment should provide not only a task-oriented
support but also an interface able to establish a social empathic relation with the user.
Several studies report successful results on how social assistive robots can be em-
ployed as interface in the assisted living domain. For instance, projects ROBOCARE
[4], Nursebot [5], Care-o-bot [6], CompaniAble [7], and Ksera [8] aim at creating as-
sistive intelligent environments for the elderly in which robots offer support to the us-
er at home. However, to be accepted and integrated in the user’s daily life, interaction
with robots must be spontaneous and natural, and to provide a friendly environment
robots must exhibit social capabilities and learn how to react according to the human
emotional state. Since speech provides a natural and intuitive way for people to inter-
act with robots, automatic emotional speech recognition will expand the possibilities
of interaction.
Emotions are expressed through various communicative signals in humans: facial
expressions [9], vocal features [10], body movements and postures [10,11], or a com-
bination of some of them [13,14,15]. In this paper we focus on speech features and
how it is possible to use them to recognize emotions in communication with humans.
Recognizing emotions in speech through several features has been a key research
issue in robotics, because by recognizing emotional factors the robot can handle social
situations. In emotional classification from speech, a multitude of different features
denoting prosodic cues have been used. Prosodic features, like pitch, loudness, speak-
ing rate, durations, pause and rhythm were proven to have strong correlations between
them, providing emotional information. In the case of the analysis of an entire seg-
ment of voice, statistical functions like mean, median, minimum, maximum, standard
deviation are applied to the fundamental frequency (F0) base contour [16]. Taking
advantage of research work in Music Information Retrieval, Mel Frequency Cepstral
Coefficients (MFCCs) are also used with great accuracy in emotion recognition [17].
These features can be used to train a classifier and the learned model can be used to
detect emotion in real-time situations.
Several classifiers have been used in this field. Each of them has advantages and
disadvantages in order to deal with the speech emotion recognition problem. The
more common group includes Hidden Markov Models (HMM) [18, 19] regarded as
the simplest dynamic Bayesian networks, artificial neural networks (ANN) [20], sup-
port vector machines (SVM) [21], k-NN [22] and Decision Trees [23].
The majority of emotion recognition systems from speech have employed a high-
dimensional speech grouped in a big vector of features. In this paper, the most com-
monly used features in several researches for capturing emotional speech characteris-
tics in time and frequency were selected. The performance of different well known
classifiers was compared in order to select the best result to predict the emotion, based
on speech emotional data.
3. Overview of the MAS
In [24] we propose an approach based on software agents able to provide what we call
Smart Services. A smart service can been defined as an integrated, interoperable and
personalized service, accessible through several interfaces available on various devi-
ces present in the environment in the optic of pervasive computing.
The objective of the proposed approach is to recognize the users goal starting from
percepts (sensors data, user actions, etc.) and provide them with a smart service that
integrate elementary services according to the situation. In order to achieve this aim,
the environment has to be able to reason on the situation of the user so as to under-
stand which are his/her needs and goals through the composition of the most appro-
priate smart service. The idea underlying our approach is the metaphor of the butler in
grand houses, who can be seen as an household affairs manager with duties of a per-
sonal assistant, able to organize the housestaff in order to satisfy the needs of the hou-
se inhabitants. To this aim, taking into account the results of a previous project, we
have developed a MAS in which the butler agent has to recognize the situation of the
user, based on interaction with Sensor Agent, in order to infer possible user’s goals.
The recognized goals are then used to select the most suitable workflow among a set
of available candidates representing a smart service. Such a selection is made by mat-
ching semantically the goals, the current situation features and the effects expected by
the execution of the workflow. Once a workflow has been selected, its actions are
executed by the effector agents.
One important feature of this architecture is the presence of an agent designed to
take care of the interaction with the users. In completely proactive approach, in fact,
users may feel a loss of control over the system actions. Therefore we adopt a semi-
automatic approach composition of services. The butler proactively propose smart ser-
vices and leaves, at the same time, the control over proposed service composition to
the user to select alternative services, to provide more preference information in order
to get a better personalization, to ask for explanation about the proposed services and
so on.
The MAS is constituited by the following classes of agents:
- Sensor Agents (SA) are used for providing information about context parameters
and features (e.g., temperature, light level, humidity, etc.) at a higher abstraction
level than sensor data.
- Butler Agent (BA) reasons on the user’s goals and devises the workflow to sati-
sfy them (see Figure 1).
- Effector Agents (EA) each appliance and device is controlled by an EA that rea-
sons on the opportunity of performing an action instead of another in the current
context.
- Interactor Agent (IA) is in charge of handling interaction with the user in order
to carry on communicative tasks. In this case the IA is embodied in the NAO Ro-
bot.
- Housekeeper Agent (HA) acts as a facilitator since it knows all the agents that
are active in the house and also the goal they are able to fulfill.
Fig. 1. The MAS architecture
3. VOCE: VOice Classifier of Emotions
Emotions can be classified using two main approaches. Discrete emotion models fo-
cus on a defined set of labels denoting emotions (e.g. anger, fear, disgust, happiness,
surprise and sadness to name the most common ones). The discrete emotion model
has the advantage of clearly distinguishing categories of emotions, however the the
labels and their number differ a lot from one model to another. By contrast, dimensio-
nal models describe the affective space within a limited amount of dimensions. For
instance in the circumplex model of emotions [25] only two dimensions are used to
represent an emotional state: the valence (from positive to negative or pleasant vs.
unpleasant) and the arousal (from high to low or aroused vs. relaxed) dimensions. In
contrast to discrete emotions, each emotion can be mapped within this space and this
model can be used to determine mixtures of different emotions because they are re-
presented by points in a space.
In VOCE we decided to adopt the dimensional model to classify emotions. There-
fore, the analysis of the prosody user’s spoken utterance is made by two classifiers:
one for recognizing the valence dimension and the other for the arousal one.
To this aim we developed a web-service called VOCE 2.0 (VOice Classifier of
Emotions ver. 2.0) that classifies the valence and arousal of the voice prosody with an
approach very similar to the one described in [26]. The major steps in speech emo-
tion recognition are audio segmentation, feature extraction and the actual classifica-
tion of the feature vectors into valence and arousal values.
VOCE can be used in two ways: offline for creating and analysing the emotional
speech corpus (Figure 2) and, being a web-service, online for tracking the affect in
voice in real-time. While the off-line version allows to build the classifier, the online
emotion recognition just outputs the recognised emotions valence and arousal and
maps the combination of these values into one of the basic emotions by providing the
emotion label.
Fig. 2. The interface of the off-line version of VOCE 2.0
Let’s see now how the classifier has been trained and how we use it in real-time.
3.1 Dataset
Albeit our approach is based on the dimensional model, since we could not found any
corpus, among the few available for Italian, in which emotions were annotated ac-
cording to their valence and arousal we used the ∈motion dataset [27]. Among the
three availble ones the ∈motion dataset has been used for EVALITA ERT challenge1
and therefore we could compare our results with other research works in this domain.
The emotional speech characteristics were extracted from the Italian subset of
∈motion contains 220 audio files corresponding to sentences for the 6 basic emotions
(joy, anger, surprise, sadness, disgust, fear) and the neutral one recorded by profes-
sional actors. In order to use the dimensional approach on this dataset we mapped
each emotion to the correspondent valence (negative, neutral and positive) and arous-
al (low, medium, high) using the approach explained in [25]. For instance, “anger” is
mapped into negative valence and high arousal, while “sadness” is mapped into nega-
tive valence and low arousal.
3.1.2 Features extraction and classification
In developing VOCE we exploited different combinations of features and several
classification algorithms. For this task we used Praat [28]. In particular, besides pitch
and energy related features, we extracted features related to the spectrum, harmonicity
and the Mel-Frequency Cepstral Coefficients (MFCCs) that is used to describe a spec-
trum frame, its first and second derivative in time are used to reflect dynamic chan-
ges.
Table 1 shows the features extracted from each audio file using Praat functions.
In order to find the best set of features we tested three conditions with several clas-
sification algorithms:
- Support Vector Machines (SVM), which offers robust classification to a very
large number of variables and small samples.
- Decision trees, that work with simple classification rules that are easy to un-
derstand. The rules represent the information in a tree based on a set of featu-
res.
- Artificial Neural Network (ANN), and in particular the Multilayer Perceptron
algorithm.
- k-Nearest Neighbors (kNN) is one of the simplest of classification algorithms
available for supervised learning. It classifies unlabeled examples based on
their similarity with examples in the training set.
The three set of features were:
- ALL: all the attributes in Table 1;
- MFCC: MFCC features only;
- No MFCC: all the features except MFCC.
1
http://www.evalita.it/2014/tasks/emotion
Feature
Description
Pitch
PitchMin
Minimum
value
PitchMed
Average
value
PitchMax
Maximum
value
PitchMinMaxDiffLog
Logarithmic
differentiation
PitchMinLog
Minimum
Logarithmic
PitchMedLog
Average
logarithmic
PitchMaxLog
Maximum
logarithmic
PitchDevSta
Standard
Deviation
PitchSlope
Slope
Energy
EnergyMin
Minimum
value
EnergyMed
Average
value
EnergyMax
Maximum
value
EnergyMinMaxDiff
Logarithmic
differentiation
EnergyDevSta
Standard
Deviation
Spectrum
SpectrumCentralMoment
Central
moment
SpectrumDevSta
Standard
Deviation
SpectrumGravityCentre
Central
tendency
SpectrumKurtosis
Degree
of
centralization
SpectrumSkewness
Degree
of
asymmetry
Harmonicity
HarmonicityMin
Minimum
value
HarmonicityMed
Average
value
HarmonicityMax
Maximum
value
HarmonicityDevSta
Standard
Deviation
MFCC
MFCCnMin
Minimum
of
nth
MFCC
MFCCnMed
Average
of
nth
MFCC
MFCCnMax
Maximum
of
nth
MFCC
MFCCnDevSta
Standard
Deviation
of
nth
MFCC
Table 1. The set of features extracted from speech for emotion recognition
From the analysis of the performance of the most commonly used algorithms for
classification starting from numeric features the most accurate one were MLP (Multi-
Layer Perceptron) and SMO (Sequential Minimal Optimization) algorithm for trai-
ning a support vector classifier in Weka2.
The accuracy was validated using a 10 Fold Cross Validation technique. A k-fold
cross-validation with k = 10 was used to make validations over the classifiers. This
technique allowed the evaluation of the model facing an unknown dataset. Results of
the classification of valence, arousal and derived emotion labels are shown in Table 2.
Results show that, for both algorithms, using the complete set of features improves
accuracy, however using only MFCC related features we get an accuracy comparable
2
http://www.cs.waikato.ac.nz/ml/weka/
with the one obtained using the all set of features. The worst setting is when MFCC is
not considered. As far as the choice of algorithm is concerned, even if MLP had a
slight better accuracy, the time to create the model and classify a vocal input is higher
(100:1). Since VOCE has to be employed in real-time classification tasks we selected
SMO.
Since the arousal dimension is related to the importance of the goal and the valen-
ce dimension is related to the achievement vs. the threatening of the goal, our speech
classifier performs well in recognizing negative states, like those related to anger, and
allows us to distinguish positive from negative attitudes. However, as expected, some
emotions are easier to recognize than others. For example, humans are much better at
recognizing anger than happiness; therefore, our results can be considered acceptable
under this view.
Features MLP SMO
ALL
Valence 70,45 71,36
Arousal 80,90 77,27
Emotion 71,36 68,63
MFCC
Valence 69,09 64,54
Arousal 80,00 75,00
Emotion 70,45 68,18
No_MFCC
Valence 64,09 55,45
Arousal 74,09 69,09
Emotion 53,18 53,18
Table 2. Accuracy of the two classifiers for valence arousal and derived emo-
tion labels.
Comparing our results with other works based on the same dataset [29] we can say
that our approach has a comparable accuracy over the set of basic emotions.
4 An Example of Application in the Context of AAL
VOCE has been used in real-time as a web service with the NAO Robot for ena-
bling emotion recognition during speech-based interaction (see Figure 3).
We have designed this architecture to endow the Aldebaran NAO robot with this
capability. The system is composed by two fundamental units: the NAO humanoid
robot and the workstation connected with NAO robot. Audio files in wav format,
recorded from 4 microphones located in the head of the NAO, are collected by the
Application Programming Interface (API) provided with NAO Software Development
Kit (SDK). Captured audio files are sent to the Speech-based Interface module in
order to allows the understaning of vocal commands and to recognize the valence and
arousal. An Automatic Speech Recognition Module (ASR) performs the first task
whereas the second task is accomplished by Voice Classifier for Emotions (VOCE)
Module. Then, the results are sent to Behavior Decision Module that choose the
appropriate behavior and send it to the robot to be executed. Communication between
the robot and the workstation has been performed using the NAOqi API.
Speech-Based
Interface
Fig. 3. Overview of the proposed system
Let us now provide an example of how the proposed approach can be integrated in
an ambient assisted living environment.
The scenario depicts a situation in which an old man lives in a house equipped
with some sensors (to gather data about the house situation) and some effectors (to
control appliances in the environments). Moreover the house is equipped with the
NAO robot acting as a natural and social interface between the user and the smart
home environment, by implementing an easy conversational access to the (digital or
physical) services of the environment.
It’s friday evening and Nicola, a 73 y.o. man, is at home alone. He has an ap-
pointment with his friends downtown to play cards like he does almost every
friday. Nicola is sitting on the bench in the living room that is equipped with
sensors and effectors for controlling appliances in the room and with the so-
cial robot that acts as a mediator interface between the environment services
and the user (see Figure 4). Nicola received a message saying that his
doughter cannot accompany him downtown and this makes him a bit angry.
Nicola calls NAO to try to find a solution.
Fig. 4 A simulation of the scenario
In the following we provide an example of the interaction.
R: ‘Hi Nicola, what can I do for you?’
Nicola: ‘Damn … (with high arousal and negative valence) I need to go downtown
to play cards with my friends and my doughter cannot come to bring me there …I
cannot miss it tonight there is a tournament !’ (with high arousal and negative va-
lence).
R: ‘Don't’ be angry for this … we will find a solution’ Do you want me to call
your daughter to ask for the permission to call a taxi to bring you there and take
you back at a certain time?”.
Nicola: ‘Yes … but you know I need a bit of assistance in walking from the car to
the bar’ (with medium arousal and negative valence).
Nicola: I will not play card with my friends tonight, I feel so lonely. (with low
arousal and negative valence).
R: ‘Oh, I’m sorry to hear that you are sad. I will ask the taxi driver to help you.
OK?
Nicola: OK.
R: the robot send a message to the daughter that accepts and then calls the taxi.
In this scenario initially the voice classifier recognizes a negative valence with a
high arousal from the prosody of the spoken utterance (Figure 2). This is interpreted
as anger and the robot besides expressing empathy (it show to understand the user’s
feelings) asks the reasons for it. When the robot understands that the user’s goal is to
go downtown it finds a workflow satisfying this goal by matching the constraints of
the situation (the daughter is not available). According to the selected workflow the
dialog goes on in order to get some information that are necessary (preconditions) to
execute some of its step (like the permission to call the taxi given from Nicola’s
daughter).
5 Conclusions and Future Work
We presented a preliminary work towards implementing a speech-based interface be-
tween a Social Robot and a user for handling interaction in a smart environment. In
our opinion, besides assisting the elderly user in performing tasks, the robot has to es-
tablish a social long-term relationship with the user so as to enforce trust and confi-
dence. The underlying idea of our work, in fact, is that the analysis of user’s spoken
utterances can be used for both issuing requests to the robot and understaning his
emotional state, which is important when the interaction happens in everyday life en-
vironments. To this aim we developed VOCE 2.0 a module to classify emotions from
features extracted from the speech signal according to the dimensional model (valence
and arousal). The recognition accuracy results are comparable with other research
works in which the same dataset was used [29]. We are aware that an improvement is
needed and to this aim we plan to collect a new dataset possibly with example of el-
derly voices, which may have different range of features from those used in the
∈motion dataset. Information on user’s emotions coupled to context and activity
recognition may give the robot the capability to infer which is or will be the most
probable user’s mood in a given context.
References
1. S. Thrun, Towards a framework for human-robot interaction, Human Computer Interaction.
19 (1&2), pp. 9-24, 2004.
2. Drygajlo, P.J. Prodanov, G. Ramel G., M. Meisser, and R. Siegwart, “On developing a
voice-enabled interface for interactive tour-guide robots”. Journal of Advanced Robotics,
vol.17, nr. 7,p.p. 599-616, 2003.
3. B. De Carolis , G. Cozzolongo , S. Pizzutilo , V. L. Plantamura, Agent-Based home
simulation and control, Proceedings of the 15th international conference on Foundations of
Intelligent Systems, May 25-28, 2005, Saratoga Springs, NY
4. Cesta, G. Cortellessa, F. Pecora and R. Rasconi, Supporting Interaction in the RoboCare
Intelligent Assistive Environment, AAAI 2007 Spring Symposium, 2007.
5. J. Pineau, M. Montemerlo, M. Pollack, N. Roy and S. Thrun, Towards Robotic Assistants
in Nursing Homes: Challenges and Results, Robotics and Autonomous Systems 42(3–4),
pp. 271–281, 2003.
6. Graf B, Hans M, Schraft RD (2004) Care-O-bot II – development of a next generation
robotics home assistant. Auton. Robots 16, 193–205.
7. CompanionAble project (2011) http://www.companionable.net
8. ksera.ieis.tue.nl
9. M. Pantic and L.J.M. Rothkrantz, “Automatic analysis of facial expressions: The state of
the art,” IEEE Trans. on Pattern Analysis and Machine Intelligence, vol. 22, no. 12, pp.
1424–1445, 2000.
10. R. Cowie and E. Douglas-Cowie, “Automatic statistical analysis of the signal and prosodic
signs of emotion in speech,” In Proc. International Conf. on Spoken Language Processing,
pp. 1989–1992, 1996. [
11. N. Bianchi-Berthouze and A. Kleinsmith, “A categorical approach to affective gesture
recognition,” Connection Science, vol. 15, no. 4, pp. 259–269, 2003.
12. G. Castellano, S.D. Villalba and A. Camurri, “Recognising Human Emotions from Body
Movement and Gesture Dynamics,” In Proc. of 2nd International Conference on Affective
Computing and Intelligent Interaction, Berlin, Heidelberg, 2007.
13. H. K. M. Meeren, C. van Heijnsbergen and B. de Gelder, “Rapid perceptual integration of
facial expression and emotional body language,” Proc. National Academy of Sciences of
the USA, vol. 102, no. 45, pp. 16518–16523, 2005. [
14. A. Metallinou, A. Katsamanis and S. Narayanan, “Tracking changes in continuous emotion
states using body language and prosodic cues,” In IEEE International Conference on
Acoustics, Speech and Signal Processing (ICASSP), pp. 2288–2291, 2011.
15. C. Busso, Z. Deng, S. Yildirim, M. Bulut, C.M. Lee, A. Kazemzaeh. S. Lee, U. Neumann
and S. Narayanan, “Analysis of Emotion Recognition using Facial Expressions, Speech and
Multimodal information,” In Proc. of ACM 6th int’l Conf. on Multimodal Interfaces
(ICMI2004), State College, PA, pp. 205–211, 2004.
16. D. Ververidis and C. Kotropoulos, “Emotional speech recognition: Resources, features, and
methods,” Speech Communication, pp. 1162– 1181, 2006
17. B. Bogert, M. Healy, and J. Tukey, “The quefrency alanysis of time series for echoes:
cepstrum, pseudo-autocovariance, cross- cepstrum, and saphe-cracking,” Proceedings of the
Symposium on Time Series Analysis, Wiley, 1963.
18. D. Le and E. M. Provost, “Emotion recognition from spontaneous speech using Hidden
Markov models with deep belief networks,” in Automatic Speech Recognition and
Understanding (ASRU), 2013 IEEE Workshop on, pp. 216–221, 2013.
19. J. Wagner, T. Vogt, and E. Andre, “A systematic comparison of different ´ HMM designs
for emotion recognition from acted and spontaneous speech,” in Proceedings of the 2nd
International Conference on Affective Computing and Intelligent Interaction (ACII),
Lisbon, Portugal, pp. 114– 125, 2007.
20. S.A. Firoz, S.A. Raj and A.P. Babu, “Automatic Emotion Recognition from Speech Using
Artificial Neural Networks with Gender-Dependent Databases,” in Advances in
Computing, Control and Telecommunication Technologies, ACT ’09, pp. 162–164, 2009.
21. C. Yu, Q. Tian, F. Cheng and S. Zhang, “Speech Emotion Recognition Using Support
Vector Machines,” in Advanced Research on Computer Science and Information
Engineering. vol. 152, G. Shen and X. Huang, Eds., ed: Springer Berlin Heidelberg, pp.
215–220, 2011.
22. M. Feraru and M. Zbancioc, “Speech emotion recognition for SROL database using
weighted KNN algorithm,” in Electronics, Computers and Artificial Intelligence (ECAI) ,
pp. 1–4, 2013.
23. C.-C. Lee, E. Mower, C. Busso, S. Lee and S. Narayanan, “Emotion recognition using a
hierarchical binary decision tree approach,” Speech Commun, vol. 53, pp. 1162–1171,
2011.
24. B. De Carolis and S. Ferilli. A multiagent system providing situation-aware services in a
smart environment. Workshop on Ambient Intelligence Infrastructures (WAmIi).
November 13, 2012, Pisa, Italy.In conjunction with International Joint Conference on
Ambient Intelligence (AmI 2012).
25. Russell, James (1980). "A circumplex model of affect". Journal of Personality and Social
Psychology 39: 1161–1178.
26. W.E. Bosma and E. André, “Exploiting Emotions to Disambiguate Dialogue Acts”, in Proc.
2004 Conference on Intelligent User Interfaces, January 13 2004, N.J. Nunes and C. Rich
(eds), Funchal, Portugal, pp. 85-92, 2004.
27. Vincenzo Galata’. 2010. Production and perception of vocal emotions: a cross-
linguistic and cross-cultural study. Ph.D. thesis, University of Calabria
28. www.praat.com
29. Antonio Origlia, Vincenzo Galatà e Bogdan Ludusan. Automatic classification of emotions
via global and local prosodic features on a multilingual emotional database. In: Speech
Prosody. 2010.