Speaker Models Research Articles

ELF- or NES-oriented pedagogy: enhancing learners’ intercultural communicative competence using a dual teaching model

Abstract This study underpins the integration of an English as a lingua franca (ELF)-oriented approach into traditional English language teaching (ELT) to develop learners’ full intercultural awareness and intercultural communicative competence (ICC). The aim is to inform prospective and senior teachers as well as practitioners about the need for adequate preparation for learners to face any speech situations involving native and non-native English speakers. Using informal classroom observations from different Italian education cycles, this study adopts a multidimensional approach to traditional ELT, combining some main ICC and ELF pedagogy tenets. Our discussion indicates that a dual teaching model, including the native English speaker (NES) model for the structural section of language teaching and learning and the ELF intercultural competent communicator model for the intercultural section, can be attained by shifting conveniently from model to model. Therefore, this study offers a fresh perspective on an extensively addressed topic by clarifying and explicitly combining the underlying connections between the main ICC and ELF tenets. This study intends to encourage teachers and practitioners to change their attitudes, perceptions, and concerns towards integrating ELF-oriented approaches into traditional ELT to tackle today’s intercultural communicative challenges outside the classroom.

Exploring Saudi EFL Students’ Attitudes Towards NES and NNES Teachers’ Accents

Various studies have examined EFL students' attitudes towards accented English. However, this topic has rarely been studied in the Saudi context. The current study examined the attitudes of Saudi EFL university students towards teachers with British or American-accented English (a native speaker model) and Arabic-accented English (a nonnative speaker model) in the language classroom. Thirteen Saudi EFL students participated in this study. An online questionnaire was administered using the matched-guise technique. The results of the study showed that EFL teachers with an American accent is the most favored by their students. Nevertheless, there was little support for this finding. Another important finding is that the examined EFL students had similar positive attitudes towards both British and Arabic accented English. This similarity might suggest that having an accent as an EFL teacher might not perceived negatively as long as the teacher has the ability to deliver information properly.

Channel and channel subband selection for speaker diarization

Speaker diarization can be considered to be one of the complex problems in speaker recognition. A reliable diarization system should be able to accurately determine the variable length utterances which a speaker contributes to multi-speaker conversations. This is a difficult problem since text-independent speaker identification and verification is yet to be improved for it to be applied reliably. While efficient speaker modelling is important for diarization, the acoustical representation of speech is the basic entity that signifies a speaker. This representation should be outstanding enough to prevent a speaker’s utterances from being lost in the acoustical congestion that is imposed by the rest of the talkers.For this purpose, it is proposed here, for the case of multiple-microphone diarization, multiple speech signals are used in the acoustic feature extraction instead of combining the signals beforehand. The reason is to make an optimal use of those signals in order to enrich the quality of the acoustical representation of the speaker. To this end, and since not all microphone signals (channels) may be desirable, two selection approaches are proposed in this work. These are, a best quality channel selection method and a novel approach for diverse channel selection. Furthermore, a novel method is proposed which retains the speech spectrum from selected least reverberated subbands of the available channels’ spectrums. A new model, referred to here as Averaged Joint Gradient (AJG), is introduced for this purpose. The proposed approach reduces the Diarization Error Rate (DER) in both of the diarization systems used in the evaluations. The first system is based on binary keys and achieves a maximum relative reduction in DER of 14%. The second one is a Gaussian Mixture Model-Bayesian Information Criterion (GMM-BIC) based system which achieves a maximum relative reduction in DER of 20%.

An Overview of the Development of Speaker Recognition Techniques for Various Applications

Speech Enhancement (SE) is a significant research issue in audio signal processing where the goal is to enhance the clarity and quality of speech signals corrupted by noise. Because of its different applications, it becomes a compelling research topic nowadays. The focus of this paper is dedicated to one of these applications which is Speaker recognition. In this paper, the fundamentals and applications of speaker recognition are discussed. A brief study on the performance and the recognition accuracy of different speaker modeling techniques has been conducted. Furthermore, while there have been several studies about the technologies used in speaker recognition, there have been few studies about the applications of speaker recognition, and none of them considers combining or linking the applications and technology in the same study. This overview demonstrates the various technologies that can be used to achieve speaker recognition applications. It aims to give a new perspective of the existing technologies uses in various applications. The paper concludes with discussions on future trends and research opportunities in this area.

4D Multimodal Speaker Model for Remote Speech Diagnosis

This paper presents a concept of a 4D multimodal speaker model (4D-MSM) for asynchronous remote speech diagnosis. Recording and archiving diagnostically significant articulation material remain an issue in computer-aided speech diagnosis. Therefore, we propose a workflow for preparing and storing reliable and easily interpretable multimodal data regarding pronunciation. According to our assumptions, data acquisition should be non-invasive, comfortable for both the patient and therapist, not interfere with the articulation process, and provide essential data of high quality. We developed and employed a dedicated device, obtaining a 15-channel spatially distributed audio signal and stable stereovision stream from two cameras focused on the lower part of the face. Our framework for data preprocessing covers digital beamforming of the multichannel audio signal, audio-video synchronization, and segmentation of words in the audio signal. Then, we use stereo data to calculate and adjust the depth map and prepare point clouds. Simultaneously, we delineate the mouth in video frames using a dedicated semi-automated segmentation algorithm. The point clouds are then textured with the camera images with superimposed mouth regions. Finally, we add the audio track to constitute the 4D-MSM. In the paper, we show the concept and detailed specification of the model and present experiments to justify the methodology. Proposed 4D-MSMs may be employed in remote speech diagnosis for objectifying and archiving diagnoses, conducting asynchronous consultations, and documenting the progress in therapy.

Text-independent Speaker Verification Using Hybrid Convolutional Neural Networks

Automatic speaker verification is an active research area for more than four decades, and the technology has gradually upgraded for real application. In this paper, a hybrid convolutional neural network (CNN) model is proposed where a combination of the 3D CNN & 2D CNN model is used for speaker verification in the text-independent scenario. For speaker verification, this novel convolutional neural network architecture was built to capture and discard speaker and non-speaker information at the same time. In the training process, the network is trained to differentiate between different identities of a speaker to establish the background model. The model development of the speaker is one of the important aspects. Most conventional techniques employed the d-vector system to create speaker models by means of an average of the features collected from the speaker utterance. Here a hybrid of convolutional neural networks model is utilized in the development and registration phases for building a speaker model. The approach suggested exceeds the existing methods of speaker verification.

DEVELOPMENT OF ACADEMIC ATTENDANCE SYSTEM USING VOICE VERIFICATION

Based on the fact that all Human Voices are different, where every individual’s voice contains unique characteristics that can be distinguished from others by using special analysis, voice verification can also be used for biometric recognition of individual. The aim of this project is to develop an automated attendance process captivating and student databases record maintaining using C++ programming and voice verification technique. The Input speech will undergo a series of voice recognition processes. The process starts with Feature extraction to obtain the attribute of voice. The features or specific characteristics for each voice are extracted by linear predictive coding technique (LPC) into feature vectors which later be used in speaker modeling. In speaker modeling stage the feature vectors of each speaker will be then processed by Gaussian Mixture Modeling (GMM) technique to generate speaker models and will be stored in a database. Verification process is done through template matching technique. Based on log-likelihood logic decision (template matching technique) with respect to the FAR (False Acceptance Rate) and FRR (False Rejection Rate), the identity of that speaker is accepted if the match is above the threshold. Once the verification result is accepted, the attendance in MySQL database for that speaker will be auto-updated.

Language education from a post-native-speakerist perspective: The case of English as an international language

Language education has traditionally been based on native-speakerism, which is defined in the present article, by simplifying Hollidays original definition, as a belief in the authority or superiority of native speakers. With the prevalence of native-speakerism, it tends to be taken for granted that non-native speakers should strive to accommodate themselves to native speaker models. However, in todays globalized world, such a conventional attitude is quickly becoming outdated. Above all, a most serious problem with native-speakerism is that it suppresses the freedom of thought and expression as fundamental human rights. Drawing on the case of English as an international language, this study aims to analyze the need for post-native-speakerism (a term attributed to Houghton and Hashimoto) in language teaching, or the need for relativizing native speaker norms for language learners. After illustrating major issues of native-speakerism, three theoretical paradigms for post-native-speakerism in global Englishes are presented, namely EIL (English as an International Language), WE (World Englishes), and ELF (English as a Lingua Franca), along with a prospect for integrating those different frameworks especially for pedagogical purposes. Then, educational objectives are summarized in terms of language skills, followed by the authors own examples of teaching methodologies and actual classroom practices in higher education. Several key concepts for EIL education emerge from these pedagogical efforts, including authenticity and critical literacy. In view of the urge to embrace diversity in the world today, this paper argues that post-native-speakerism is of vital importance as it allows language users to express their true selves in global communication. While many of the discussions in the present article stem from linguacultural and educational situations in Japan, it is assumed that the insights should often be applicable also to other Expanding Circle, or EFL (English as a Foreign Language), countries such as Russia and China.

Physics-based sound radiation estimation from multiple speakers by combined lumped parameter and reduced-order finite element modeling

In this paper a virtual sensor is derived that can perform full field estimation of the sound field radiating from a number of loudspeakers in the time domain. The virtual sensor scheme contains two parts: Firstly, a time-domain physics-based numerical system level model that includes both the electro-mechanical behavior of the loudspeaker and the resulting wave propagation. Secondly, a Kalman filter in which the numerical model and a limited set of microphone measurements are combined to perform full field estimation. The loudspeaker dynamics are captured in a lumped parameter model and the resulting wave propagation is captured in a finite element model. To reduce the calculational complexity of the system level model, Krylov subspace based reduced order modeling is applied on the finite element model, leading to a significant reduction of the number of states. The system level model is brought to a discrete time state–space format, which allows for efficient implementation in the Kalman filter. The method is validated using two setups that use both vented and closed box loudspeakers, leading to improved estimation results as compared to pure simulation results, thus showing the merits of the virtual sensor. Furthermore, the importance of including the lumped parameter speaker model in the system level model is shown by comparing the estimation results with and without the inclusion of the lumped parameter model.

Classification of emotions and evaluation of customer satisfaction from speech in real world acoustic environments

This paper focuses on finding suitable features to robustly recognize emotions and evaluate customer satisfaction (CS) from speech in real acoustic scenarios. The classification of emotions is based on standard and well-known corpora and the evaluation of CS is based on recordings of real opinions given by customers about the received service during phone calls with call-center agents. The feature sets considered in this study include two speaker models, namely x-vectors and i-vectors, and also the well known feature set introduced in the Interspeech 2010 Paralinguistics Challenge (I2010PC). Additionally, we introduce the use of phonation, articulation and prosody features extracted with the DisVoice framework as alternative feature sets to robustly model emotions and CS from speech. The results indicate that the I2010PC feature set is the best approach to classify emotions in the standard databases typically used in the literature. When considering the recordings collected in the call-center, without any control over the acoustic conditions, the best results are those that include our articulation features. The I2010PC feature set includes 1584 measures while the articulation approach only includes 488 measures. We think that the proposed approach is more suitable for real-world applications where the acoustic conditions are not controlled and also it is potentially more convenient for industrial applications.

Accentron: Foreign accent conversion to arbitrary non-native speakers using zero-shot learning

Foreign accent conversion (FAC) aims to create a new voice that has the voice identity of a given second-language (L2) speaker but with a native (L1) accent. Previous FAC approaches usually require training a separate model for each L2 speaker and, more importantly, generally require considerable speech data from each L2 speaker for training. To address these limitations, we propose Accentron, an approach that can generate accent-converted speech for arbitrary L2 speakers unseen during training. In the proposed approach, we first train a speaker-independent acoustic model on L1 corpora to extract bottleneck features that represent the linguistic content of utterances. Then, we develop a speaker encoder and an accent encoder to generate embedding vectors for the desired voice identity (L2 speaker’s) and accent (L1 accent), respectively. Lastly, we use a sequence-to-sequence model to transform bottleneck-features to Mel-spectrograms, conditioned on the L2 speaker embedding and the L1 accent embedding. We conducted experiments on the L2-ARCTIC corpus under two testing conditions: the standard FAC setting where test L2 speakers were seen during training, and a zero-shot FAC setting where test L2 speakers were unseen during training. Accentron achieves over 27% relative improvement in accentedness ratings compared to two state-of-the-art FAC systems in the standard FAC setting. More importantly, our results show that Accentron generalizes to the zero-shot FAC setting with no performance loss. Therefore, in practical use scenarios (e.g., computer-assisted pronunciation training software), Accentron can effectively avoid the need to adapt or retrain the model, which significantly reduces computations and the users’ waiting time.

Modeling speaker-specific vocal tract kinematics from gestural scores

The theory of Task Dynamics provides a method of predicting articulatory kinematics from a discrete phonologically-relevant representation (“gestural score”). However, because the implementations of that model (e.g., Nam et al., 2004) have generally used a simplified articulatory geometry (Mermelstein et al., 1981) whose forward model (from articulator to constriction coordinates) can be analytically derived, quantitative predictions of the model for individual human vocal tracts have not been possible. Recently, methods of deriving individual speaker forward models from real-time MRI data have been developed (Sorensen et al., 2019). This has further allowed development of task dynamic models for individual speakers, which make quantitative predictions. Thus far, however, these models (Alexander et al., 2019) could only synthesize limited types of utterances due to their inability to model temporally overlapping gestures. An updated implementation is presented, which can accommodate overlapping gestures and incorporates an optimization loop to improve the fit of modeled articulatory trajectories to the observed ones. Using an analysis-by-synthesis approach, the updated implementation can be utilized: (1) to refine the hypothesized speaker-general gestural parameters (target, stiffness) for individual speakers; (2) to test different degrees of temporal overlapping among multiple gestures such as a CCVC syllable. [Work supported by NSF, Grant 1908865.]

PROMINENCE AND MEANING SELECTION IN GHANAIAN ENGLISH DISCOURSE: TOWARDS A CATEGORIZATION OF THE NEW ENGLISHES

All models of intonation recognize that speakers in any communication context select prominence based on their communicative intents. Consequently, every prominent syllable bears a specific communicative function. However, while prominence selection appears fixed in inner circle Englishes, the same cannot be said of outer (OC) and expanding circle (EC) Englishes. Results from research clearly suggest that it is not always possible for OC and EC speakers to select prominence to indicate meaning selection. This study reports findings on the selection of prominence in Ghanaian English in relation to Brazil’s Discourse Intonation model. Data consisting of 6 hours of English conversations from 100 Ghanaians were analyzed using Brazil’s model. Results show that just as in other OC English varieties, Ghanaians assign prominence to individual syllables with communicative intent, although their selections may not always coincide with the functions proposed in Brazil’s model. Based on these results, it is argued that rather than being viewed as deficiencies, phonological choices such as prominence selection should be regarded as common to OC Englishes. The study, therefore, argues for an adoption of a model that describes the intonation of outer circle Englishes in relation to their contexts and not on a native speaker model.
 
 <p> </p><p><strong> Article visualizations:</strong></p><p><img src="/-counters-/edu_01/0856/a.php" alt="Hit counter" /></p>

Two-Year-Old Children Expect Native, but Not Foreign Speakers to Use the Same Tool for the Same Purpose.

Previous research has already demonstrated that even very young children are sensitive to language cues and learn differently from native and foreign speaker models. A possible explanation for this phenomenon suggests that spoken language is a sign of someone's cultural background and in this sense demonstrates the person's culture specific knowledge. The aim of the present study was to investigate what children think about native and foreign speakers' behavior in a domain that is typically regulated by cultural norms (tool usage), specifically whether they expect group members to act alike or not. In a violation of expectation paradigm, two-year-old toddlers first watched a video on which a native and a foreign speaker person used different tools for achieving the same goal. In the test phase a new native speaker model appeared and selected one of the previously seen tools for the same goal as it was used before. Results indicated that toddlers were surprised if the native speaker model had chosen the tool that had previously been used by the foreign speaker. In Experiment 2, the familiarization phase was exactly the same as in Experiment 1, but during the test phase, the model spoke a foreign language. Results, in this case, showed no significant differences between looking times. These experiments suggest that two-year-olds expect native (but not foreign) speakers to use the same tool for the same goals. As tool usage is a fundamental element of cultural knowledge, we propose that this pattern of results suggest that children expect native speakers to possess shared cultural knowledge at least in the domain of artifacts.

Informativity, topicality, and speech cost: comparing models of speakers’ choices of referring expressions

ABSTRACT This study formalizes and compares two major hypotheses in speakers’ choices of referring expressions: the topicality model that chooses a form based on the topicality of the referent, and the rational model that chooses a form based on the informativity of the form and its speech cost. Simulations suggest that both the topicality of the referent and the informativity of the word are important to consider in speakers’ choices of reference forms, while a speech cost metric that prefers shorter forms may not be.

The time course of speaker-specific language processing

Listeners are sensitive to a speaker's individual language use and generate expectations for particular speakers. It is unclear, however, how such expectations affect online language processing. In the present EEG study, we presented thirty-two participants with auditory sentence stimuli of two speakers. Speakers differed in their use of two particular syntactic structures, easy subject-initial SOV structures and more difficult object-initial OSV structures. One speaker, the SOV-Speaker, had a high proportion of SOV sentences (75%) and a low proportion of OSV sentences (25%), and vice-versa for the OSV-Speaker. Participants were exposed to the speakers' individual language use in a training session followed by a test session on the consecutive day. ERP-results show that early stages of sentence processing are driven by syntactic processing only and are unaffected by speaker-specific expectations. In a late stage, however, an interaction between speaker and syntax information was observed. For the SOV-Speaker condition, the classical P600-effect reflected the effort of processing difficult and unexpected sentence structures. For the OSV-Speaker condition, both structures elicited different responses on frontal electrodes, possibly indexing effort to switch from a local speaker model to a global model of language use. Overall, the study identifies distinct neural mechanisms related to speaker-specific expectations.

Current English Speaker Models in Senior High School Classrooms

Current English Speaker Models in Senior High School Classrooms

Efficient Approach of Automatic Speech Emotion Recognition (ASR) Using Mutual Information

Speech is a significant quality for distinguishing a person in daily human to human interaction/ communication. Like other biometric measures, such as face, iris and fingerprints, voice can therefore be used as a biometric measure for perceiving or identifying the person. Speaker recognition is almost the same as a kind of voice recognition in which the speaker is identified from the expression instead of the message. Automatic Speaker Recognition (ASR) is the way to identify people who rely on highlights that are omitted from speech expressions. Speech signals are awesome correspondence media that constantly pass on rich and useful knowledge, such as a speaker's feeling, sexual orientation, complement, and other interesting attributes. In any speaker identification, the essential task is to delete helpful highlights and allow for significant examples of speaker models.
 Hypothetical description, organization of the full state of feeling and the modalities of articulation of feeling are added. A SER framework is developed to conduct this investigation, in view of different classifiers and different techniques for extracting highlights. In this work various machine learning algorithms are investigated to identify decision boundary in feature space of audio signals. Moreover novelty of this art lies in improving the performance of classical machine learning algorithms using information theory based feature selection methods. The higher accuracy retrieved is 96 percent using Random forest algorithm incorporated with Joint Mutual information feature selection method.

SPEESH: speech-based mobile application for dysarthric speech recognition

There are numerous automatic speech recognition applications available in market, but these applications are quite challenging to be used for individual with speech disability. Due to this demand, SPEESH, a speech recognition system is developed for individuals with dysarthria disorders. The main objective of this project is to study the requirement of proposed system and characteristics of individuals with dysarthria. Interview with speech therapists in Malaysia has been conducted on the phone to elicit the requirements for the system. Based on the interview, we acquired numerous common words that are used by dysarthric patients. These findings are crucial in designing the application to ease the user’s experience. The system consists of two phases on how it works; text dependent and text independent. First, in a text dependent phase, the system will learn to recognize specific words from the speaker and register them into database. In the next phase, speaker model will now consist characteristics of voice of the person and it can identify what one is saying. SPEESH application have been designed to be able to translate words from individuals with dysarthria or any speech disorders. All the functionalities and features are all developed in a way such that SPEESH will work as a communication platform between dysarthria individuals and their surroundings.

Analysis of transition cost and model parameters in speaker diarization for meetings

There has been little work in the literature on the speaker diarization of meetings with multiple distance microphones since the publications in 2012 related to the last National Institute of Standards (NIST) Rich Transcription Evaluation Campaign in 2009 (RT09). Lately, the Second DIHARD Challenge Evaluation has also covered diarization at dinner party meetings that include multiple distant microphones. Dinner party meetings are somehow harder than office meetings because their participants can move freely around the room. In this paper, we studied some of the algorithms on speaker diarization for meetings with multiple distant microphones for the NIST Rich Transcription Evaluation Campaign in 2007 (RT07) and RT09 and provide definite and clear improvements. On the one hand, little or no care has been taken to the problem of penalizing or favoring transitions between speakers other than proposing a minimum duration of a speaker turn or calculating the speakers’ probabilities using Variational Bayes (VB). We have studied this issue and determined that a transition penalty term is needed that should be independent both of the number of active speakers and the minimum duration of speaker turns. On the other hand, the determination of a method to automatically select the right number of parameters is crucial in developing good models for speakers. Previous studies have proposed the dynamic selection of the number of parameters based on the duration of the speaker’s speech with a mixed performance when tested at one distant microphone meetings or multiple distant microphones meetings. In this paper, we propose a new method that takes into account both the duration of speaker’s speech to determine a minimum number of parameters, and the question of overfitting issue to determine a maximum number of them, also taking into account the computation time in order to reduce it.We have carried out experiments to support our findings, and we have been able to improve our baseline speaker error rate with multiple distant-microphone meetings. Both methods achieve improved performance over the baseline. The first method obtains a 21.6% decrease in relative speaker error for the development set and a 4.6% decrease in relative speaker error for the test set (RT09). The second method obtains a 46.47% decrease in relative speaker error for the development set and a 17.54% decrease in relative speaker error for the test set. Both methods complement each other, and when they are applied in combination, we obtain a 47.2% decrease in relative speaker error for the development set and a 22.02% decrease in relative speaker error for the test set.The performance obtained with our proposal is outstanding in some subsets of the development test such as the NIST RT07 and among the best for RT09 using our proposed simple modifications. Furthermore, with our algorithm we obtain gains in computation time without jeopardizing performance. Results with a different publicly available database, augmented multiparty interaction (AMI) obtains a 28.44% decrease in relative speaker error confirming the validity of our methods. Preliminary experiments with a single stream (mfcc) endorse the validity of our findings. Comparisons with an x-vector system deliver superior performance of our system on unseen test data.