hash stringlengths 32 32 | doc_id stringlengths 7 13 | section stringlengths 3 121 | content stringlengths 0 2.2M |
|---|---|---|---|
0501906322d59d0b9af5ed9b13e45aa2 | 103 327 | 5.1.6.4 Service Discovery Phase | Atomic services are discovered automatically when the sensor is discovered, because each sensor is used for a process which is considered as atomic service. For example, if a sensor is used for heart rate measurements, the atomic service shall be the Heart rate service. Simple services may be discovered base on SWRL ru... |
0501906322d59d0b9af5ed9b13e45aa2 | 103 327 | 5.1.6.5 Service Processing Phase | The service processing phase depends on the service process description where input/output parameters, precondition, post-condition, method and effects are described. For each service, the developer shall define a Web Service/agent in order to execute the adequate process (Service Processing Agent). But, before executi... |
0501906322d59d0b9af5ed9b13e45aa2 | 103 327 | 5.1.6.6 Network Management | Few existing architectures provided the possibility for an interoperable solution for WBAN remote management, regardless the network, data link and physical layers used in the deployed WBAN. One of the added values of the SmartBAN Reference Architecture depicted in the present document is the handling of this possibili... |
0501906322d59d0b9af5ed9b13e45aa2 | 103 327 | 5.1.7 Summary | In summary, the SmartBAN Reference Architecture depicted in the present document permits the automatic nodes discovery, services discovery, services matching, remote monitoring, and automatic notification, with the minimum user and developer interaction. It masks the heterogeneity between different nodes and communicat... |
0501906322d59d0b9af5ed9b13e45aa2 | 103 327 | 5.2.1 Validation use case: elderly at home monitoring | Smart home for elderly fall monitoring system has been the use case that has been selected for the validation of the SmartBAN Reference Architecture depicted in the present document. The system consists of: 1) BAN, the body area network containing: - The HR sensor (in our use case the H7Polar Sensor); Note that more bi... |
0501906322d59d0b9af5ed9b13e45aa2 | 103 327 | 5.2.2 Tests and results | The applications were tested on Android smartPhones (Android 4.4.2 KITKAT-API 19 and Android 5.0.1-API 21 Lollipop after upgrade and Android 4.1.2 Jelly Bean- API 16). One patient per WSN is first considered. The application is installed on two different patients' mobile. For each patient, one relative was added. Then,... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 1 Scope | The present document specifies Classification of and Requirements for Emotion Detectors for telecommunications and the assessment of their performance and uncertainties. |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 2 References | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 2.1 Normative references | References are either specific (identified by date of publication and/or edition number or version number) or non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the referenced document (including any amendments) applies. Referenced documents which a... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 2.2 Informative references | References are either specific (identified by date of publication and/or edition number or version number) or non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the referenced document (including any amendments) applies. NOTE: While any hyperlinks i... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 3 Definitions, symbols and abbreviations | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 3.1 Definitions | For the purposes of the present document, the following terms and definitions apply: curragator: muscle of the upper face which assists in expressing an emotion diphone: unit of speech made up of two simple speech sounds known as phones electroglottograph: device for measuring how much electricity flows across the lary... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 3.2 Symbols | For the purposes of the present document, the following symbols apply: BWn bandwidths of formants E( ) Emotion vector ESRM percentage of Emotional Samples in Resulting Material ESSM percentage of Emotional Samples in Source Material F F-measure F0 fundamental frequency of the human voice fn percentage of false negative... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 3.3 Abbreviations | For the purposes of the present document, the following abbreviations apply: ACR Absolute Category Rating AI Artificial Intelligence ANN Artificial Neural Networks ASR Automatic Speaker Recognition ATR Advanced Telecommunications Research BAWE British Academic Written English BTW By The Way ATR Lab Advanced Training & ... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4 Emotion detectors for written text | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.1 Introduction | In recent years, recognition of emotions, sentiment analysis (SA), affect analysis and opinion mining have become a very popular topic for researches, the reason being that emotional state can influence the ways people behave. For example, the analysis of the emotional state is widely used by e-shops to analyse custome... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.2 Overview of emotion detectors | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.2.1 Introduction | Most of the emotion detectors and emotion recognition started with facial analysis [i.1] to [i.3], then continued to voice and developed to textual emotion recognition as well. Emotional detector is a powerful tool as it can be used for improving customer experience, removing inappropriate posts from social networks an... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.2.2 Current approaches | Whether the goal is emotion recognition from image, voice or text, the means to achieve those are the same. Modern machine learning techniques made it possible to classify these emotions from images, but also from voice. Advanced signal processing algorithms are employed to do that. A summary of selected papers used in... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.2.3 Emotion detector description | The emotion detectors have many interpretations, yet most of them can be described as the input textual data, which is processed in an algorithm to produce emotional output. The top level view of such system is shown in figure 1. These systems have many different realizations for the algorithm, which may be using metho... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.3 Input | In this clause the main inputs are discussed, their sources, and requirements and recommendations for them. For the EmoDet model 2 types of input information were selected: text and meta-data, which can be derived from video or audio recordings. For the first type of information (text), inputs can be arranged into 4 gr... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.4 Linguistic resources | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.4.1 Introduction | People use words to express happiness, frustration, despair, etc. but sometimes it is more complex to understand the underlying meaning of the words, for example when using irony or sarcasm. Another example is using idioms which hold a very specific meaning. This is a very demanding task. It becomes even harder when em... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.4.2 Overview of resources | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.4.2.1 Description | Generally, authors use the term linguistic resource for any ontology, corpus, or lexicon. Ontology is a way of describing things by their properties with a structured and specific manner. In different field ontology may have a slightly different meaning. From the top point of view ontology is a way of organizing object... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.4.2.2 WordNet | The most frequently used linguistic resource in the field of NLP for emotion recognition in text seems to be WordNet. Although it is described as a large lexical database of English, some consider it ontology as well. Nouns, verbs and other classes of words are grouped into sets of cognitive synonyms (synsets), each ex... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.4.2.3 Suggested Upper Merged Ontology (SUMO) | The Suggested Upper Merged Ontology is free and owned by IEEE. Its domain ontologies form the largest formal public ontology in existence today. It has about 25 000 terms and about 80 000 axioms when all domain ontologies are combined. Main target areas are research and applications in search, linguistics and reasoning... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.4.2.5 Open Mind Common Sense (OMCS) | In its nature is similar to Cyc™. It is also a knowledge base collecting fundamental human knowledge, which is taken as granted by humans. The knowledge was gained since 2000 from OMCS project. The project leveraged the power of social interactions on the Internet and people were filling in information in a questionnai... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.4.2.6 ConceptNet | It represents knowledge from OMCS as a directed graph. Nodes in the graph are concepts and labelled edges assertions. |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.4.2.7 Other databases | Other examples of linguistic resources used in NLP applications today are Ren-CECps (corpus in Chinese language) British Academic Written English (BAWE), International Survey on Emotion Antecedents and Reactions (ISEAR), Multi Perspective Question Answering (MPQA) and Micro-WordNet-Opinion (Micro-WNOp) Some linguistic ... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.4.3 Annotation | The data in a knowledge base is usually annotated by people to avoid any false annotations. Unfortunately this process is lengthy and expensive. Human annotators are slow and to annotate a large database of words it is necessary to employ many annotators. More importantly - to achieve a high level of precision, it is n... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.5 Emotion models | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.5.1 Introduction | The effort to analyse human emotions started a long time ago and it is still an ongoing process. The first person who tried to deal with emotions was Cicero. He determined four basic emotion categories: "metus" (fear), "aegritudo" (pain), "libido" (lust), "laetitia" (pleasure). Discrete categories with labels reflectin... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.5.2 Categorical emotion classification | Categories which are used in the research of textual emotional recognition are based on the idea of discrete emotion theory. This theory argues that there is only a limited set of emotions which a person can experience. To this category belong the following: • Positive arousal or negative arousal. • Discrete emotional ... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.5.3 Dimensional emotional classification | Dimensional classification methods introduce a different concept. Emotions are presented as a point or a region within a two-dimensional or multi-dimensional space. Therefore, they are not subject of assignment to one category, but to many variables. These methods extend the categorical classification by adding a scale... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6 Algorithms | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.1 Introduction | Basically, emotion-detection approaches can be classified into 3 following types: • keyword based/lexical-based; • learning based; • hybrid. ETSI ETSI TS 103 296 V1.1.1 (2016-08) 29 |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.2 Keyword based approach | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.2.1 Description | In this observation the keyword based approach was the most used technique. The idea of this approach is to detect emotions in the text at the basic word level and it is very effective mostly for analysing emotion-bearing words and for simple sentences with clearly expressed emotions (detailed information about inputs ... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.2.2 Advantages | The main distinguishing factors of this type of approach are: • no need of machine-learning process; • easily implementable. |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.2.3 Disadvantages | However, this approach has very significant limitations: • First of all, it has a weakness regarding the recognition of negative sentences and sentences with double negation. For example, the algorithm will have problems to determine the polarity of the sentence "The new episode of Star Wars was expected to be the exhi... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.2.4 Implementation | It has been stated that the most of the keyword based approaches have similar constructions of implementation. For this method the following sequence of steps is recommended in order to perform a high-accuracy emotion detection [i.41]: • Lexicon of emotion expressing words has to be created. For this purpose linguistic... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.2.5 Related works and results | Two types of results are presented in tables of related works. The first type is F-score which is calculated using 2 parameters called precision and recall. Precision and recall are measured using 4 parameters: • tp - the percentage of true positive detected samples; • tn - the percentage of true negative detected samp... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3 Learning based approaches | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.1 Descriptions | Learning based approaches generally have a training set which is used to train the classifier. Then the classifier is used to determine either emotion directly for a word or a more complex structure of classifiers. There are many types of learning based approaches. From the analysed papers it has been stated that the f... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.2 Support Vector Machines | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.2.1 Description | A commonly used and highly effective approach for identification of emotional states is the Support Vector Machines (SVM) binary classification technique, which is considered as a state of the art method. The idea of this learning based algorithm is that it uses training examples, which contain information about their ... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.2.2 Advantages | The most important advantages worth to mentioning are: • Ability of SVM to recognize hand-written text. • High accuracy of prediction. • Correct performance even when training samples are not fully accurate. |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.2.3 Disadvantages | • The SVM is able to classify data only into 2 categories. However, implementing improvement algorithms can reduce the problem of multitasking [i.6]. • It is required that Input data are fully labelled, otherwise a helping algorithm called Support Vector Clustering (SVC) should be implemented [i.43]. • Compared to othe... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.2.4 Related works and results | Table 5 represents some related works using SVM approach and obtained results. Table 5: Related works using SVM algorithm and obtained results Reference Year Algorithm Language F-score in % Accuracy in % [i.42] 2002 SVM with features based on unigrams English N/A 82,9 [i.11] 2007 SVM with General Inquirer, WordNet Affe... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.3 Naïve Bayes Classifier | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.3.1 Description | Another effective and simple learning based approach is the Naïve Bayes Classifier (NBC) algorithm. There are various types of NBC depending on representation of input text. However, the multinomial Naïve Bayes model is more accurate than other event models [i.46]. In this model, the frequencies of specific words' occu... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.3.2 Advantages | • A very meaningful advantage of NBC is that it can be trained using only a small amount of training data. • Has minimal consumption of CPU and memory [i.48]. • Fast training and classifying. • Can deal with both real and discrete data. • Inappropriate features included in training data do not have impact on its perfor... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.3.3 Disadvantages | This classifier's disadvantage is that: • Despite the statement that features in class are independent of each other, its independence assumptions are very naïve, which means that among variables typically exist some dependencies. |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.3.4 Related works and results | Table 6 includes works related to the NBC and obtained results. Table 6: Works related to Naïve Bayes Classifier and obtained results Reference Year Algorithm Language F-score in % Accuracy in % [i.42] 2002 NBC with features based on unigrams English N/A 78,7 [i.51] 2012 NBC and Naïve Search Arabic N/A about 85 [i.50] ... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.4 Hidden Markov Model | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.4.1 Description | The Hidden Markov Model (HMM) is also simple and frequently used learning based algorithm for emotion detection. Basically, HMM is a technique which is able to distribute classes over sequence of observations. In other words, it is possible to anticipate the future of the process as the process encapsulates all the imp... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.4.2 Advantages | It is needed to observe only the previous state of the process. |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.4.3 Disadvantages | The main disadvantage is that it is very hard and inefficient to interpret the HMM model with big number of states and objects and over-fitting can occur. For this purpose, extensions such as Factorial HMMs, Tree structured HMMs and Switching State space models can be implemented. |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.3.4.4 Related works and results | Table 7 includes works related to HMM algorithm and obtained results. Table 7: Works related to Hidden Markov Model and obtained results Reference Year Algorithm Language F-score in % Accuracy in % [i.54] 2003 Continuous HMM German and English N/A 77,8 [i.15] 2012 A High-order HMM with Viterbi algorithm English 35,3 N/... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.4 Hybrid approaches | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.4.1 Description | This type of approaches is combination of already mentioned approach types. It is able to detect emotions based on detected keywords, learned patterns and other additional information from various dictionaries and thesauri [i.56]. Another advantage is that this approach can minimize different resources' integration com... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.6.4.2 Related works and results | Table 8: Works related to hybrid approaches and obtained results Reference Year Algorithm Language F-score in % Accuracy in % [i.57] 2004 Hybrid SVM (PMI/Osgood and Lemmas), 100 folds English N/A 89 [i.9] 2013 Keyword-spotting method and rule-based method English 76.97 N/A [i.58] 2013 Multinomial NBC with greedy search... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.7 Output | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.7.1 Output description | The output block is responsible for post processing of data produced by the algorithm and delivery of results to the user. The post processing is done in terms of organizing the information in a human or machine readable form, which can be later saved to a file. Delivery to the user means generating the output file and... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.7.2 Practical Examples | A very convenient way of expressing emotional background of words proved to be colour coding. The use of colours and other visualizations can leverage the emotional information in many different contexts. Valkanas [i.12] takes advantage of the six Ekman's basic emotions coded directly to a colour space. ETSI ETSI TS 10... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 4.8 Final remarks on textual emotion detectors | This clause has analysed document current works, approaches and ideas in order to make understood the notion of Emotion Detector, considering all the inputs, knowledge bases, emotion models, algorithms and outputs. Clause 4.2 introduces ways of textual emotion recognition and suggests what all the approaches have in co... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.1 Preface | Emotions frequently influence the speech behaviour. In particular, a change of prosodic parameters, i.e. fundamental frequency of human voice F0, formants, duration of speech unit and distribution of energy in a spectrum, is a part of emotions. Emotions depend on the language carrying a message and the social backgroun... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.2 Introduction | The study and development of systems and devices that can recognize interpret, process and simulate human affects are parts of affective computing. It is an interdisciplinary field. In fact, emotions involve different components, such as subjective experience, expressive behaviour, psychophysiological changes and instr... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.3 Basic information about speech emotions | Robert Plutchik defines emotions as an inferred complex sequence of reactions to a stimulus including cognitive evaluations, subjective changes, autonomic and neural arousal, impulses to action and behaviour designed to have an effect upon the stimulus that initiated the complex sequence. Four major traditions that hav... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.4 Language | Aneta Pavlenko from TESOL at the College of Education, Temple University, Philadelphia is the author of many publications about emotions and bilingualism resp. multilingualism [i.88]. In bilingual societies, people frequently switch between languages, particularly in emotional situations. Researchers from the Universit... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5 Existing Corpora | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.1 Preface | The collection of emotion corpora or databases is a basic necessity. Many authors divide emotions into primary and secondary or into active and passive. Its number is different for each particular group. The following text and tables provide more detailed information about existing databases. The overview is shown in t... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.2 Introduction | The collection of emotional speech databases is a necessary condition for emotion speech pre-processing. A large source of emotional databases information created so far is presented in [i.77]. Authors Dimitrios Ververidis and Constantine Kotropoulos reviewed 32 emotional speech databases. Each database consists of a c... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.3 English speech emotion databases | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.3.1 Preface | Most databases concentrate on English language (16 databases - separately or in combination with other languages). |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.3.2 Database 1 | R. Cowie and E. Cowie [i.98], [i.99] and [i.100] constructed this database at the Queen's University of Belfast. It contains emotional speech in five emotional states: anger, sadness, happiness, fear and neutral (see clause 5.3 for details). The speakers are 40 volunteers (20 female, 20 male) aged between 18 years and ... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.3.3 Database 2 | R. Cowie and M. Schroder constructed Belfast Natural Database at Queen's University [i.91]. The database is designed to sample genuine emotional states and to allow exploration of the emotions through time. Two kinds of recordings took place. One was recorded in studio and the other from TV programs. A total of 239 cli... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.3.4 Database 3 | R. Cole and his assistants at the University of Colorado recorded Kids' Audio Speech Corpus NSF/ITR Reading Project [i.101]. The aim of the project was to collect sufficient audio and video data from kids in order to enable the development of auditory and visual recognition systems, which enable face-to-face conversati... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.3.5 Database 4 | M. Liberman, Kelly Davis and Murray Grossman at the University of Pennsylvania constructed the Emotional Prosody Speech and Transcripts [i.102]. The database consists of 9 hours of speech data. It contains speech in 15 emotional categories, such as hot anger, cold anger, panic-anxiety, despair, sadness, elation, happin... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.3.6 Database 5 | J. Hansen at the University of Colorado Boulder has constructed in 1999 an emotional speech corpus the so called SUSAS (Speech Under Simulated and Actual Stress). The database contains voice from 32 speakers (13 female and 19 male) with ages ranging from 22 to 76 years old. In addition, four military helicopter pilots ... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.3.7 Database 6 | C. Pereira at Macquire University constructed this database [i.103]. The database consists of 40 sentences uttered by two actors in five emotional categories. There are two repetitions of these 40 utterances, thus creating 80 presentations altogether. In the study, 31 normal hearing subjects (18 men and 13 women) rated... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.3.8 Database 7 | M. Edgington at BT Labs, UK collected this emotional speech database for training a voice synthesizer [i.105]. One professional male actor was employed. The database contains speech in six emotional categories, such as anger, fear, sadness, boredom, happiness and neutral (see clause 5.3 for details). A laryngograph was... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.3.9 Database 8 | T. S. Polzin and A. H. Waibel at the Carnegie Mellon University constructed this emotional speech database [i.106]. The database consists of emotional speech in five emotional categories. The corpus was comprised of 291 word tokens per emotion per speaker. The sentence length varies from 2 to 12 words. These sentences ... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.3.10 Database 9 | V. Petrushin at the Centre for Strategic Technology Research, Accenture constructed this database in order to train neural networks in speech recognition [i.107]. It is divided into two studies. The first study deals with a corpus of 700 short utterances expressed by 30 professional actors. The corpus contains emotiona... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.3.11 Database 10 | R. Fernandez at MIT labs constructed emotional speech this database [i.108]. The subjects are drivers who were asked to sum up two numbers while driving a car. The questions produced by a speech synthesizer and the sum of the numbers was less than 100. The two independent variables in this experiment were the driving s... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.4 German speech emotion databases | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.4.1 Preface | Most databases concentrate on German language (9 databases - separately or in combination with other languages). |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.4.2 Database 11 | The Verbmobil database was recorded at the University of Hamburg [i.109] and [i.110]. It is non-neutral emotion oriented, because it contains mainly anger and dissatisfaction. The database contains voice from 58 native German speakers, between 19 and 61 years old (29 male, 29 female), while they were speaking to an emu... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.4.3 Database 12 | The SmartKom Multimodal Corpus was constructed at the Institute for Phonetics and Oral Communication in Munich, with national funding [i.109] and [i.111]. The database consists of "Wizard of OZ" dialogues (like Verbmobil database) in German and English. The database contains multiple audio channels (which follow speake... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.4.4 Database 13 | W. F. Sendlmeier et al. at the Technical University of Berlin constructed another emotional speech database [i.112] and [i.113]. The database consists of emotional speech in seven emotion categories. Each one of the 10 professional actors expresses 10 words and five sentences in all the emotional categories. The corpus... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.4.5 Database 14 | K. Alter at the Max-Planck-Institute of Cognitive Neuroscience constructed an emotional speech database for medical purposes [i.114]. Electroencephalogram (EEG) recordings were also taken. The aim of the project was to relate emotions, which are recognized from speech with a location in the human brain. A trained femal... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.4.6 Database 15 | K. Scherer at the University of Geneva constructed this database [i.111]. The purpose of his study was to bring into light the differences in emotional speech perception between people from different countries. Results showed that the Indonesian people understand the emotions in speech in a different way. Four German p... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.4.7 Database 16 | B. Wendt and H. Scheich at Leibniz Institute of Neurobiology constructed the Magdeburger Prosodie Korpus [i.115]. The aim was to construct a brain map of emotions. The database contains emotional speech in six emotional categories. The database contains also the word accentuation, word length, speed rate, abstraction/c... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.4.8 Database 17 | M. Schroeder and M. Grice recorded a database of diphones that can be used for emotional speech synthesis [i.116]. One male speaker of standard German produced a full German diphone set for each of three degrees of vocal effort: "soft", "modal" and "loud". Four experts verified the vocal effort, the pitch constancy and... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.4.9 Database 18 | M. Schroder has constructed an emotional speech database, which consists of "affect bursts" [i.117]. His study shows that affect bursts, presented without context, can convey a clearly identifiable emotional meaning. Professionals selected the affect bursts from the German literature. Altogether, the database comprises... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.5 Japanese speech emotion databases | |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.5.1 Preface | Three databases focus on Japanese language. Two of them are collected at the ATR Laboratories, one of them at the Kyoto Institute of Technology. |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.5.2 Database 19 | R. Nakatsu et al. at the ATR Laboratories constructed an emotional speech database in Japanese [i.118]. The database contains speech in eight emotion categories. The project employed 100 native speakers (50 male and 50 female) and one professional radio speaker. The professional speaker was told to read 100 neutral wor... |
0b4f01803e8c54dd6b19e65cbd20b96a | 103 296 | 5.5.5.3 Database 20 | Y. Niimi et al. at the Kyoto Institute of Technology, Matsugasaki, Japan developed another emotion speech database in Japanese [i.119]. It consists of VCV (Vowel Consonant Vowel) segments for each of the three emotion speech categories, such as anger, sadness and joy. These VCVs can generate any accent pattern of Japan... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.