المؤلفون: Blaauw, Merlijn

المساهمون: University/Department: Universitat Pompeu Fabra. Departament de Tecnologies de la Informació i les Comunicacions

مرشدي الرسالة: Gómez Gutiérrez, Emilia, Bonada, Jordi

المصدر: TDX (Tesis Doctorals en Xarxa)

مصطلحات موضوعية: Singing synthesis, Deep learning, Autoregressive modeling, Fast inference, Reduced effort voice creation, Síntesis de canto, Aprendizaje profundo, Modelado autorregresivo, Inferencia rápida, Creación de voces con esfuerzo reducido, Síntesi de cant, Aprenentatge profund, Modelat autorregressiu, Inferència ràpida, Creació de veus amb esforç reduït

الوصف: Singing synthesis has seen a notable surge in popularity in the last decade and a half. Music producers use this technology as an instrument, there is an audience for music with synthetic vocals, and an entire range of cultural phenomena surrounding singing synthesis has emerged. At the time of starting this work, the prevailing approaches for singing synthesis were concatenative synthesis on the one hand, and hidden Markov model synthesis on the other. Concatenative synthesis was state of the art in terms of quality, but lacked flexibility due to being based on signal processing, heuristics and carefully prepared data. By contrast, hidden Markov model synthesis is based on data-driven machine learning, which brings a certain degree of flexibility, but was never able to match the sound quality of concatenative synthesis. At the same time, the field of text-to-speech started to shift towards powerful new deep learning models that have shown to be able to combine high-quality results with a high degree of flexibility. In this dissertation, we try to answer whether similar models can also live up to this potential for singing synthesis. We also try to answer whether these approaches allow fast and stable synthesis, qualities important for many real-world applications. Finally, we try to answer whether the flexibility that the deep learning approaches offer allows creating new voices with smaller amounts of data, and less effort (time, expert knowledge), which is a notable bottleneck in older approaches. To this end, we propose a number of singing synthesis models, and evaluate them, principally through listening tests. The first part of this dissertation focuses on modeling timbre, via autoregressive and non-autoregressive models. The second part focuses on improving data efficiency through voice cloning, reducing the voice creation effort by using a sequence-to-sequence mechanism that requires fewer annotations, and a semi-supervised model which combines supervised pre-training with unsupervised training of a new target voice. Through our experiments, we show deep learning methods can not only outperform the previous state of the art, they can also allow for a significantly reduced voice creation effort. With our work on these elemental problems in singing synthesis, we hope that future research can advance the field further by focusing on topics such as expression, user control and non-modal voice qualities.

الوصف (مترجم): La síntesis de canto ha visto un aumento notable en popularidad en la última década y media. Los productores de música utilizan esta tecnología como instrumento, existe una audiencia para la música con voces sintéticas y ha surgido toda una gama de fenómenos culturales en torno a la síntesis del canto. Al momento de comenzar este trabajo, los enfoques principales para la síntesis de canto eran la síntesis concatenativa por un lado y la síntesis basada en modelos ocultos de Márkov por el otro. La síntesis concatenativa era estado del arte en términos de calidad, pero carecía de flexibilidad debido a estar basado en el procesamiento de señales, heurísticas y datos cuidadosamente preparados. Por el contrario, la síntesis basada en modelos ocultos de Márkov se basa en el aprendizaje automático usando datos, lo que brinda cierto grado de flexibilidad, pero nunca pudo igualar la calidad de sonido de la síntesis concatenativa. Al mismo tiempo, el campo de «text-to-speech» comenzó a cambiar hacia nuevos y poderosos modelos de «deep learning» que han demostrado ser capaces de combinar resultados de alta calidad con un alto grado de flexibilidad. En esta tesis, tratamos de responder si modelos similares también pueden estar a la altura de este potencial para la síntesis de canto. También tratamos de responder si estos enfoques permiten una síntesis rápida y estable, cualidades importantes para muchas aplicaciones del mundo real. Finalmente, tratamos de responder si la flexibilidad que ofrece el «deep learning» permite crear nuevas voces con cantidades más pequeñas de datos y menos esfuerzo (tiempo, conocimiento experto), lo cual es un cuello de botella importante enlos enfoques previos. Para ello, proponemos una serie de modelos de síntesis de canto y los evaluamos, principalmente a través de pruebas de escucha. La primera parte de esta tesis se centra en el modelado del timbre, a través de modelos autorregresivos y no autorregresivos. La segunda parte se centra en mejorar la eficiencia de los datos a través de la clonación de voz, reduciendo el esfuerzo de creación de voz mediante el uso de un mecanismo «sequence-to-sequence» que requiere menos anotaciones y un modelo semisupervisado que combina un entrenamiento previo supervisado con un entrenamiento no supervisado de la nueva voz deseada. A través de nuestros experimentos, mostramos que los métodos de «deep learning» no solo pueden superar el estado del arte anterior, sino que también pueden permitir un esfuerzo de creación de voz significativamente reducido. Con nuestro trabajo sobre estos problemas elementales en la síntesis del canto, esperamos que la investigación futura pueda avanzar más en el campo centrándose en temas como la expresión, el control del usuario y las cualidades de voz no modales.
La síntesi del cant ha experimentat un notable augment de popularitat en l’última dècada i mitja. Els productors musicals utilitzen aquesta tecnologia com a instrument, existeix un públic interessat en la música amb veus sintètiques i ha sorgit tota una sèrie de fenòmens culturals al voltant de la síntesi del cant. En el moment d’iniciar aquest treball, els enfocaments predominants per a la síntesi del cant eren la síntesi concatenativa d’una banda i la síntesi basada en els models ocults de Màrkov de l’altra. La síntesi concatenativa era l’estat de l’art en termes de qualitat, però mancada de flexibilitat perquè es basa en el processament del senyal, l’heurística i les dades curosament preparades. Per contra, la síntesi basada en models ocults de Màrkov es fonamenta en l’aprenentatge automàtic a partir de dades, que aporta un cert grau de flexibilitat, però que mai no ha igualat la qualitat de la síntesi concatenativa. Al mateix temps, el camp «text-to-speech» va començar a canviar cap a nous models potents de «deep learning» que han demostrat ser capaços de combinar resultats d’alta qualitat amb un alt grau de flexibilitat. En aquesta tesi, intentem respondre si models similars també poden estar a l’altura d’aquest potencial per a la síntesi del cant. També intentem respondre si aquests enfocaments permeten una síntesi ràpida i estable, qualitats importants per a moltes aplicacions del món real. Finalment, intentem respondre si la flexibilitat que ofereix el «deep learning» permet crear noves veus emprant menors quantitats de dades i menys esforç (temps, coneixement expert), que és un dels grans coll d’ampolla dels enfocaments anteriors. Per a això, proposem una sèrie de models de síntesi de cant, i els avaluem, principalment mitjançant proves auditives. La primera part d’aquesta tesi se centra en la modelització del timbre, mitjançant models autoregressius i no autoregressius. La segona part se centra a millorar l’eficiència de les dades mitjançant la clonació de veu, reduint l’esforç de creació de veu mitjançant l’ús d’un mecanisme «sequence-to-sequence» que requereix menys anotacions, i un model semisupervisat que combina un entrenament previ supervisat amb un entrenament no supervisat de la nova veu desitjada. A través dels nostres experiments, mostrem que els mètodes de «deep learning» no només poden superar l’estat de l’art anterior, sinó que alhora permeten un esforç de creació de veu significativament reduït. Amb el nostre treball sobre aquests problemes elementals en la síntesi del cant, esperem que properes investigacions puguin avançar encara més centrant-se en temes com l’expressió, el control d’usuari i les qualitats de veu no modals.
Programa de doctorat en Tecnologies de la Informació i les Comunicacions

وصف الملف: application/pdf

الوصول الحر: http://hdl.handle.net/10803/675120Test

المؤلفون: Sundblad, Göran¹ (AUTHOR) goran.sundblad@slu.se, de Groote, Annica² (AUTHOR), Lundquist, Peter² (AUTHOR), Larsson, Mattias³ (AUTHOR), Linderfalk, Rasmus⁴ (AUTHOR), Larson, Fredrik⁵ (AUTHOR)

المصدر: Fisheries Management & Ecology. Apr2024, Vol. 31 Issue 2, p1-11. 11p.

مصطلحات موضوعية: *CRAYFISH, *FISHING catch effort, *FISH traps, *FISHING, *FISHERY management

مصطلحات جغرافية: SWEDEN

مستخلص: Estimates of recreational fishing effort and catch are needed to inform fisheries management, but such estimates can be challenging to obtain. We combined a zig‐zag line transect survey of buoys, a fisheries enforcement survey and a drone survey to estimate recreational effort in the open‐access trap fishery for signal crayfish (Pacifastacus leniusculus) in the sixth largest lake in Europe, Lake Vättern, Sweden. Using supporting variables such as size of nearby ports and bathymetry, our model‐assisted estimate of total effort was 6400 (±4300 CI) buoys during the three‐weekend season. Using the frequency of buoys to traps as a constant, the total number of crayfish traps was estimated to be 25,500 (±17,000 CI). The sampling design and methods presented are useful for ecologists and resource managers to design future multi‐source surveys of recreational trap fishing effort. [ABSTRACT FROM AUTHOR]

المؤلفون: Stratoudakis, Yorgos¹ (AUTHOR) yorgos@ipma.pt, Vieira, Manuel^2,3 (AUTHOR), Marques, João Pedro³ (AUTHOR), Amorim, Maria Clara P.^2,3 (AUTHOR), Fonseca, Paulo J.^2,4 (AUTHOR), Quintella, Bernardo R.^2,3 (AUTHOR)

المصدر: Reviews in Fish Biology & Fisheries. Mar2024, Vol. 34 Issue 1, p491-510. 20p.

مصطلحات موضوعية: *FISHERY management, *FISHING catch effort, *SOUND pressure, *SMALL-scale fisheries, *ESTUARIES

مصطلحات جغرافية: PORTUGAL

مستخلص: Passive acoustic monitoring (PAM) is useful for monitoring vocal fish but has had so far limited application in fisheries management. Here, four years (2016–2019) of concurrent daily catch and effort fishery data in Portugal and species-specific vocal activity in the Tagus estuary are compared to describe biological and small-scale fishery dynamics for a large sciaenid fish, the meagre (Argyrosomus regius), that aggregates to spawn. Consistent patterns in seasonality of acoustic and fisheries variables indicate that most fishing takes place within the Tagus estuary in spring and summer months, when higher vocal activity related to spawning aggregations is detected in the PAM station. Good fit of statistical models shows that PAM (sound pressure level in the third-octave band with centre frequency at 500 kHz during dusk) and PAM-supported variables (mean weekly catch per first sale transaction) can provide useful surveillance indicators to improve local management. Signs of overexploitation and hyperstability are detected and communicated to the estuarine fishing communities with the aim to initiate an adaptive local management cycle. The approach can be relevant for fisheries targeting other vocal fish that seasonally aggregate and face similar threats of overexploitation. [ABSTRACT FROM AUTHOR]

المؤلفون: Robertson, Matthew D.¹ (AUTHOR) matthew.robertson@mi.mun.ca, Midway, Stephen R.² (AUTHOR), Embke, Holly S.³ (AUTHOR), Kaz, Anna L.² (AUTHOR), Lang, Mitchel⁴ (AUTHOR), Paukert, Craig⁵ (AUTHOR), Sievert, Nicholas A.⁵ (AUTHOR), Wszola, Lyndsie⁶ (AUTHOR), Lynch, Abigail J.⁷ (AUTHOR)

المصدر: Fisheries Management & Ecology. Feb2024, Vol. 31 Issue 1, p1-12. 12p.

مصطلحات موضوعية: *FISHING catch effort, *FISHING surveys, *SOCIOECONOMIC factors, *BODIES of water, *FISHERIES

مصطلحات جغرافية: UNITED States

مستخلص: Recreational fisheries represent a socially, ecologically, and economically significant component of global fisheries. The U.S. Inland Creel and Angler Survey Catalog (CreelCat) database includes inland recreational fisheries survey data across the United States to facilitate large‐scale analyses. However, because survey methods differ, a statistical method capable of integrating these surveys is necessary to assess patterns and relationships across regions. Here, we developed a hierarchical generalized linear mixed modeling approach to estimate the relationship between daily recreational fisheries catch and effort based on waterbody, socio‐economic, and ecological covariates. We applied this approach to CreelCat data on lentic waterbodies and found that recreational fisheries catch and effort were non‐linearly related (i.e., catch per unit of effort declined as effort increased), where effort varied regionally and by waterbody area, median county age, and distance to nearest primary road. This modeling approach could be used to inform data‐poor regions or waterbodies, make comparisons across spatial scales, and, with the inclusion of socio‐economic and ecological factors, inform management techniques in an era of shifting demographics and landscapes. [ABSTRACT FROM AUTHOR]

المؤلفون: Beaudoin, Richard, author

الإتاحة: https://doi.org/10.1093/oso/9780197659281.001.0001Test

المؤلفون: Kekenusa, John Socrates¹ (AUTHOR) johnskekenusa@unsrat.ac.id, Rondonuwu, Sendy Beatrix² (AUTHOR) sendy43@gmail.com, Paendong, Marline Sofiana¹ (AUTHOR) marlinepaendong@unsrat.ac.id

المصدر: AIP Conference Proceedings. 2023, Vol. 2877 Issue 1, p1-9. 9p.

مصطلحات موضوعية: *FISHING catch effort, *MACKERELS, *RENEWABLE natural resources

مستخلص: Mackerel scad (Decapterus spp) needs to be managed well because even as a renewable natural resources, but can undergo depletion or extinction. One of the approach in the management of the resources is by modelling. The analysis was performed aiming to get the best estimated for the surplus production model to determine the maximum sustainable yields (MSY), utilization level, and effort level of mackerel scad. The data of catch and fishing effort mackerel scad collected from the Marine and Fisheries Service of the Manado Regency and the North Sulawesi Province. Best Surplus Production Model, which is used to assess the potential of mackerel scad is Fox Model. Optimal effort (EMSY) of 1,000 trips per year, with catches of optimal CMSY 3,037.394 tons per year. The effort level for 2013 is 266.5 %, which shows the inefficiency of effort, the utilization level of 71.08 %, showing the catches is lower than optimum. [ABSTRACT FROM AUTHOR]

المؤلفون: Oijaku, Obinna C., Ezenwafor, Ebuka Christian, Osarenkhoe, Aihie, 1960

المصدر: International Journal of Technology Marketing. 18(1):113-135

مصطلحات موضوعية: task-technology fit, TTF, UTAUT, performance expectancy, effort expectancy, financial technologies, Fintech, adoption intention, consumers’ intentions, emerging country context, Intelligent industri, Intelligent Industry

الوصف: Previous studies show that although financial technologies (Fintech) bridge the financial inclusion gap, its rate of adoption is low. This study aims to develop understand of factors that influence consumers’ intentions to adopt financial technologies (Fintech) for savings, loans, and investment by testing a model that integrates the unified theory of acceptance and usage of technology (UTAUT) and task-technology fit (TTF) model. Data was collected from a cross-section of 324 respondents online. Structural equation modelling via smart PLS 3.0 was used to test the hypothesised relationship. The result showed that the TTF and UTAUT integrative model is robust and adequately explained Fintech adoption. Performance expectancy and social influence significantly affected behavioural intentions, and effort expectancy significantly predicts performance expectancy. Interestingly, TTF predicts use behaviour but not significant on adoption intention. Finally, task characteristics strongly predict effort expectancy and performance expectancy. Focusing on user perceptions of the technology and neglecting the effect of the task technology fit, as commonly done in extant literature, may be not enough. Thus, this study fills this gap and integrates both UTAUT and TTF to facilitate understanding of illuminate factors that influence consumers’ intentions to adopt financial technologies in an emerging country context.

وصف الملف: electronic

الوصول الحر: https://urn.kb.se/resolve?urn=urn:nbn:se:hig:diva-43185Test
https://doi.org/10.1504/IJTMKT.2024.135674Test
https://hig.diva-portal.org/smash/get/diva2:1807515/FULLTEXT01.pdfTest

المؤلفون: Sousa-Ribeiro, Marta, Stengård, Johanna, 1972, Leineweber, Constanze, 1973, Bernhard-Oettel, Claudia

المصدر: Work, Aging and Retirement. 10(3):225-240

مصطلحات موضوعية: preferred retirement age trajectories, latent class growth curve modeling, work ability, health, effort-reward imbalance, psykologi, Psychology

الوصف: Preferred retirement age (PRA) is one key dimension when studying retirement decision-making. However, little is known concerning how PRA develops over the late career years. This study used a person-centered approach to longitudinally investigate trajectories of PRA and how they differ in self-rated health, perceived work ability, and effort–reward imbalance (ERI) at baseline levels and over 6 years. The study used data from four waves (2010, 2012, 2014, and 2016) of the Swedish Longitudinal Occupational Survey of Health. The sample consisted of 1,510 individuals aged 50–55 in 2010, who answered to the questionnaire for those in paid work (including self-employment) at the baseline and at least one of the following waves. Results from the latent class growth curve modeling show both within- and between-person variability in PRA over the 6-year span. We found four distinct trajectories, which differed both at the baseline levels and in the patterns of change in PRA: “C1: normative, relatively stable PRA” (42% of all participants); “C2: considerably early, increasing PRA” (6% of the participants); “C3: late, relatively stable PRA” (4% of the participants); and “C4: early, increasing PRA” (49% of the participants). Participants revealed a clear preference for retirement before the age of 65. Trajectories comprising earlier PRA showed poorer self-rated health, poorer work ability, and higher levels of ERI at the baseline and over time. The findings reinforce the importance of healthy work environments that promote work ability and facilitate a balance between efforts and rewards for encouraging longer working lives.

وصف الملف: print

الوصول الحر: https://urn.kb.se/resolve?urn=urn:nbn:se:su:diva-216308Test
https://doi.org/10.1093/workar/waad006Test

المؤلفون: Anselme, Patrick, Oeksuez, Fatma, Okur, Nurdem, Pusch, Roland, Güntürkün, Onur

المصدر: International Journal of Comparative Psychology. 36(1)

مصطلحات موضوعية: Effort, motivation, Pecking, Unpredictability, Matching

الوصف: When well-known food resources are running out, animals extinguish their foraging behavior in that food patch and increasingly work for reward-related information to decrease outcome uncertainty. In the absence of such information, a potentially successful strategy consists of spending more time and effort searching for profitable locations—a phenomenon known from extinction learning experiments conducted in conventional conditioning chambers. Here, we tested this hypothesis by means of a semi-natural “foraging board” allowing pigeons to move and look for food items inside perforated holes. The holes could be covered with a slit plastic tape, hiding the food items they contained while making them accessible to pigeons. Our goal was to determine how pigeons forage on hidden food items in an area associated with uncertainty (one in three holes baited, on average) when visible or hidden food items were available in an adjacent area associated with certainty (each hole baited). The number of food items was equivalent in both areas. We expected longer time spent and more pecks given in the uncertain vs. certain area with the food items visible in the certain area, as well as longer time spent and more pecks given per visit in the uncertain vs. certain area with the food items either visible or hidden in the certain area. Our results confirm these predictions.

وصف الملف: application/pdf

الوصول الحر: https://escholarship.org/uc/item/5z5566mxTest

المؤلفون: Kral, Marijke, Gorissen, Pierre

المساهمون: Academie Educatie, HAN University of Applied Sciences

مصطلحات موضوعية: collaborative effort between vocational and higher education

الوصف: Personalized Learning with ICT: A Collaborative Effort between Vocational and Higher Education (Case study) EAPRIL 2023 Belfast

الوصول الحر: https://surfsharekit.nl/public/6adfef06-443c-4679-8fde-fc49756f66cdTest
https://surfsharekit.nl/objectstore/0ca98a82-450b-4969-9bae-26e75956fbd4Test

الإتاحة: http://www.hbo-kennisbank.nl/en/page/hborecord.view/?uploadId=sharekit_han:oai:surfsharekit.nl:6adfef06-443c-4679-8fde-fc49756f66cdTest