Spectral Representations of Nonmodal Phonation

We propose a nonparallel data-driven emotional speech conversion method. It enables the transfer of emotion-related characteristics of a speech signal while preserving the speaker's identity and linguistic content. Most existing approaches require parallel data and time alignment, which is not available in most real applications. We achieve nonparallel training based on an unsupervised style transfer technique, which learns a translation model between two distributions instead of a deterministic one-to-one mapping between paired examples. The conversion model consists of an encoder and a decoder for each emotion domain. We assume that the speech signal can be decomposed into an emotion-invariant content code and an emotion-related style code in latent space. Emotion conversion is performed by extracting and recombining the content code of the source speech and the style code of the target emotion. We tested our method on a nonparallel corpora with four emotions. Both subjective and objective evaluations show the effectiveness of our approach.

This paper focuses on the variety of Italian spoken in Bozen (South Tyrol, Italy) by Italian L1 speakers. Due to the peculiar multilingual situation of South Tyrol, Italian speakers are concentrated in the big towns of this area (e.g. Bozen), whereas the majority of Tyrolean people speak German or German dialects. Leaving apart the interaction problems between the two speaking groups, this work focuses on Italian community settled in Bozen by emphasizing the composition of their linguistic repertoire, and the differences among the districts of the town, also in terms of linguistic identity. For this purpose, the few data actually at disposal on this linguistic variety will be analyzed, in order to highlight the presence and use of Italo-Romance dialects by Italian speakers. A field work on this variety has never been carried on before, thus all previous statements were based only on aprioristic assumption (i.e. they were not data-driven). Without data of sort, scholars assumed the absence of spoken Romance dialects in the variety of Italian spoken in Bozen, basing on the belief that Italian speakers were not originally settled in this town. Arguing against those previous assumptions, this paper will show how dialects constitute a central part of Italian linguistic identity in the multilingual setting of South Tyrol, as it emerges from a collection of expressions collected in the so-called “Slang of Bozen”, and from speakers’ linguistic perceptions recorded during face-to-face interviews. The results of this study point out that a large number of dialects are used inside the family, and sometimes also for informal communication, whereas some words and expressions are more largely used as part of a we-code characteristic of the Italians of Bozen.

Voice conversion is a challenging task which transforms the voice characteristics of a source speaker to a target speaker without changing linguistic content. Recently, there have been many works on many-to-many Voice Conversion (VC) based on Variational Autoencoder (VAEs) achieving good results, however, these methods lack the ability to disentangle speaker identity and linguistic content to achieve good performance on unseen speaker scenarios. In this paper, we propose a new method based on feature disentanglement to tackle many to many voice conversion. The method has the capability to disentangle speaker identity and linguistic content from utterances, it can convert from many source speakers to many target speakers with a single autoencoder network. Moreover, it naturally deals with the unseen target speaker scenarios. We perform both objective and subjective evaluations to show the competitive performance of our proposed method compared with other state-of-the-art models in terms of naturalness and target speaker similarity.

It has become an integral part of language teacher education to engage teacher candidates in examining their own social and linguistic identities and dispositions and reflecting on how differences in power may lead them to rationalize inequalities and make assumptions about the sociopolitical dimensions of language use, particularly in regards to the identities of native and non-native English speakers. Such exploration of identities requires both deliberate guidance and thoughtful reflection. For these reasons, this chapter discusses activity that educates preservice teachers about English varieties and engages them in (re)examining their own perspectives and positionality while considering intersection of social (e.g. ethnicity, race, social class) and relational identities (e.g. student, teacher, peer) and language use. The activity contributes to transformative approaches that foster critical awareness of the complexity of sociolinguistic realities in preservice teacher education practices.

This article examines how racial and linguistic identities are constructed on the Australian reality TV show Border Security. Based on an analysis of 108 episodes of the show involving 253 border force officers and 128 passengers, we explore how the hegemonic Australian identity of the White native speaker of English is constructed on the show. Officers are represented as a relatively uniform group of heroes devoted to protecting Australia’s national security. Simultaneously, most of them look white and sound like native speakers of Australian English. In contrast to the officers, passengers, as their antagonists, do not have a predominant racial or linguistic profile. They are represented as highly diverse. What unites them is not any racial or linguistic profile but that they represent a security risk. Threat thus comes to be mapped onto diversity. The show’s schema of heroes and antagonists invites the audience to identify with the heroes. By identifying with the White-English heroes, the audience also comes to take on their power of judgment over its diverse linguistic and racial Others. The analysis shows how the White-English identity bundle is constructed as the authoritative and legitimate position of the judging knower. The article’s main contribution is to show how the raciolinguistic construct of the White-English complex is made hegemonic in a diverse society officially committed to multiculturalism.

The ideal goal of voice conversion is to convert the source speaker’s speech to sound naturally like the target speaker while maintaining the linguistic content and the prosody of the source speech. However, current approaches are insufficient to achieve comprehensive source prosody transfer and target speaker timbre preservation in the converted speech, and the quality of the converted speech is also unsatisfied due to the mismatch between the acoustic model and the vocoder. In this paper, we leverage the recent advances in information perturbation and propose a fully end-to-end approach to conduct high-quality voice conversion. We first adopt information perturbation to remove speaker-related information in the source speech to disentangle speaker timbre and linguistic content and thus the linguistic information is subsequently modeled by a content encoder. To better transfer the prosody of the source speech to the target, we particularly introduce a speaker-related pitch encoder which can maintain the general pitch pattern of the source speaker while flexibly modifying the pitch intensity of the generated speech. Finally, one-shot voice conversion is set up through continuous speaker space modeling. Experimental results indicate that the proposed end-to-end approach significantly outperforms the state-of-the-art models in terms of intelligibility, naturalness, and speaker similarity.

The dependence of stimulated Brillouin backscatter upon pulse amplitudes, timing, target material and geometry from a prepulse formed plasma is found. The applicability of these results to backscatter from a structured continuous pulse is discussed.

Background/Objective: Bilaterally implanted cochlear implant (CI) users do not consistently have access to interaural time differences (ITDs). ITDs are crucial for restoring the ability to localize sounds and understand speech in noisy environments. Lack of access to ITDs is partly due to lack of communication between clinical processors across the ears and partly because processors must use relatively high rates of stimulation to encode envelope information. Speech understanding is best at higher stimulation rates, but sensitivity to ITDs in the timing of pulses is best at low stimulation rates. Methods: We implemented a practical "mixed rate" strategy that encodes ITD information using a low stimulation rate on some channels and speech information using high rates on the remaining channels. The strategy was tested using a bilaterally synchronized research processor, the CCi-MOBILE. Nine bilaterally implanted CI users were tested on speech understanding and were asked to judge the location of a sound based on ITDs encoded using this strategy. Results: Performance was similar in both tasks between the control strategy and the new strategy. Conclusions: We discuss the benefits and drawbacks of the sound coding strategy and provide guidelines for utilizing synchronized processors for developing strategies.

This thesis describes experimental observations on the parametric amplification, in single crystal Yttrium Iron Garnet of elastic, magnetoelastic modes, by application of a pump pulse of twice the frequency of the L-band micro-wave signal pulse, and the excitation of a delayed pulse by the application of two time separated microwave pulses of the frequency. Details of experiments performed to determine the internal magnetostatic field profile of a magnetised rectangular cross-section Yttruim Iron Garnet rod are derived. The experimental internal fields are compared with those derived theoretically. It was found to be possible to reduce the attenuation of elastic echoes by parametric amplification. Details are given of saturation levels, attenuation and timing of the pump pulse. In the magnetolelastic regime details are given of the variation of gain wit, timing of the pump pulse, pump power, signal power orientation of the rod to the external field. Magnetostatic propagating mode, the amplitude of the parametrically excited echo is plotted against, time of application of the pump pulse, pump power and signal power. Magnetostatic standing wave mode, the amplitude of the parametrically excited echo is plotted against, time application of the pump pulse, pump power and signal power. Magnetostatic wave mode echo, parametrically excited by the application of two time separated pulses of the same frequency, plots are shown of gain relative to the first pulse with variation of, second pulse amplitude, the first pulse amplitude and time between the application of the two pulses. Parametric amplification of magnetostatic modes due to subsidiary absorbtion of the pump frequency. Plots of attenuation of the delayed pulse against signal power, pump power and time of application of the pump pulse are given. This is found to be a two port phenomenon. At L-band subsidiary absorbtion is coincident with the main absorbtion, due to the low signal frequency used, 12KMc/s, the ends of the rod were not saturated.

Variational auto-encoder (VAE) is an effective neural network architecture to disentangle a speech utterance into speaker identity and linguistic content latent embeddings, then generate an utterance for a target speaker from that of a source speaker. This is possible by concatenating the identity embedding of the target speaker and the content embedding of the source speaker uttering a desired sentence. In this work, we propose to improve VAE models with self-attention and structural regularization (RGSM). Specifically, we found a suitable location of VAE’s decoder to add a self-attention layer for incorporating non-local information in generating a converted utterance and hiding the source speaker’s identity. We applied relaxed groupwise splitting method (RGSM) to regularize network weights and remarkably enhance generalization performance. In experiments of zero-shot many-to-many voice conversion task on VCTK data set, with the self-attention layer and relaxed group-wise splitting method, our model achieves a gain of speaker classification accuracy on unseen speakers by 28.3% while slightly improved conversion voice quality in terms of MOSNet scores. Our encouraging findings point to future research on integrating more variety of attention structures in VAE framework while controlling model size and overfitting for advancing zero-shot many-to-many voice conversions <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> . <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> The work was partially supported by NSF grants DMS-1854434 and DMS-1952644 at UC Irvine.

For logging pulsed neutron tube with diameter less than 25 mm, a smaller diameter Penning ion source with a cylindrical plasma volume of Φ8 mm × 9 mm was designed to investigate the pulse characteristics of anode current. The pulse time and amplitude characteristics of anode current as a function of deuterium pressure and anode voltage were studied experimentally. The preferred working parameters for this smaller diameter Penning ion source were suggested, then obtained the pulse amplitude of anode current in the range of 1 mA∼ 3.5 mA, with the shortest pulse delay time of anode current less than 5 μs. The extract current was running stably with the extract efficiency of 22%∼ 27% for neutron tube application.

Green synthesis of new ionic liquid and its electrochemical determination at some detergents and cosmetics samples using differential pulse polarography

&lt;div class="htmlview paragraph"&gt;In the course of our studies on pulsed jet combustion (PJC) a series of tests were carried out to determine the performance of multi-stream systems. This included multi-jet efflux, as well as dual-generator aggregates. The essential performance indicator of a PJC system is the pressure transducer record, yielding, in particular, data on the rise time and amplitude of the pressure pulse, as well as the rate of pressure rise. Concomitantly with pressure measurements, a sequence of schlieren records were obtained to reveal the salient features of the fluid mechanical phenomena of the PJC process. Elucidated thereby are the effects of entrainment induced by large scale vortex structures of turbulent jet plumes, and their decisive role in governing the process of combustion, as influenced by their particular configuration. Our tests demonstrated the capability of multiple stream PJC systems to shorten the rise time and increase the amplitude of the pressure pulse, both effects of obvious potential benefit to engine performance.&lt;/div&gt;

Since recovery time of the RF coil is long at low field MRI, the rising and the ring-down times of the square pulse are also long, which means the applied sinc pulse can easily be distorted from the changing amplitude. However, both the rising time and ring-down time can be calculated using Q-factor. Using this information, an RF square pulse were compensated by appending two square pulses before and after the RF pulse. The durations of these RF square pulses were calculated using the Q-factor. Since the amplitude of the sinc pulse changes continuously, a series of square pulses were applied to apply sinc pulse to the coil. The minimum number of square pulses and the amplitude of the square pulses were calculated. It was successfully demonstrated that the sinc pulse can be compensated using a series of square pulses. The more number of square pulses were used, the smoother sinc pulse was applied to the RF coil. The Q-factor was experimentally calculated from the ring-down time of a signal induced in a sniffer loop which was connected to an oscilloscope. The resulting Q-factor was then used to calculate both the duration and amplitude of the square pulses for compensation. Echo trains were also acquired in an inhomogeneous B0 field using the compensated RF pulses. In order to enhance the SNR of the echo trains, a pre-polarization pulse was added to the CPMG spin echo sequence. The SNRs of the echo signal acquired using compensated pulses were compared with those of signal obtained with uncompensated pulses and showed significant improvements of 61.1% and 51.5% for the square and sinc shaped pulses respectively.

A family of digitally controlled devices is constructed for functional electrical stimulation in which each module is an hermetically sealed glass capsule that is small enough to be injected through the lumen of a hypodermic needle. The overall design and component characteristics of microstimulators that receive power and command signals by inductive coupling from a single, externally worn coil are described. Each device stores power between stimulus pulses by charging an electrolytic capacitor formed by its two electrodes, made of sintered, anodised tantalum and electrochemically activated iridium, respectively. Externally, a highly efficient class E amplifier provides power and digitally encoded command signals to control the amplitude, duration and timing of pulses from up to 256 such microstimulators.