Segmental durations predicted with a neural network

Teixeira, João Paulo; Freitas, Diamantino

doi:10.21437/eurospeech.2003-91

Cited by 23 publications

(5 citation statements)

References 13 publications

(12 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Preliminary work with 15 subjects gave an average score of 3.2 and 3.1 for the ACs predicted with labeled and predicted FCs, respectively, against 4.6 for the original stimulus. The ensemble usage of the whole prosody system, for durations [8] and F0, achieves a score of 3.0. The general scores achieved of 3 are at the "fair" level in a MOS scale.…”

Section: Discussionmentioning

confidence: 99%

“…It is the last part of a prosody system that has been developed for text-to-speech synthesis of EP. This system consists of a specific model for prediction of the segmental durations [8] and two other models for prediction of F0 contours based on Fujisaki's FC and AC. The complete prosody system produces contours that modulate the speech that is to be produced from the given text.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Prediction of accent commands for the Fujisaki intonation model

Teixeira,

Freitas,

Fujisaki

2004

Speech Prosody 2004

View full text Add to dashboard Cite

This paper presents a model to predict the accent commands (henceforth ACs) of the Fujisaki Model for the F0 contour, being known the phrase commands (henceforth FCs). Accent commands are associated with syllables. For each syllable, an artificial neural network (ANN) decides, with an accuracy of 89.4% whether there will be an associated AC or not. For syllables with associated AC, the amplitude, Aa, the onset time anticipation, T1a, and the offset time anticipation, T2a, are predicted by additional ANNs, with resulting linear correlation coefficient of 0.602, 0.743 and 0.650, respectively. The features used for each ANN are presented and discussed. Finally a comparison between target and predicted F0 contour is presented.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Prediction of accent commands for the Fujisaki intonation model

Teixeira,

Freitas,

Fujisaki

2004

Speech Prosody 2004

View full text Add to dashboard Cite

show abstract

“…This training, supervised or unsupervised, is based on the presentation of examples, and simulates a systematic learning process by determining the difference between the response given by the network and the expected behavior. The experience of the network is stored by the synaptic weights between neurons and its performance is evaluated, for example, by the ability to generalize behaviors, recognize patterns, fix errors or execute predictions [13][14][15].…”

Section: Multilayer Perceptron -Artificial Neural Networkmentioning

confidence: 99%

A COVID-19 time series forecasting model based on MLP ANN

Borghi

Zakordonets

Teixeira

2021

Procedia Computer Science

View full text Add to dashboard Cite

With the accelerated spread of COVID-19 worldwide and its potentially fatal effects on human health, the development of a tool that effectively describes and predicts the number of infected cases and deaths over time becomes relevant. This makes it possible for administrative sectors and the population itself to become aware and act more precisely. In this work, a machine learning model based on the multilayer Perceptron artificial neural network structure was used, which effectively predicts the behavior of the series mentioned in up to six days. The model, which is trained with data from 30 countries together in a 20-day context, is assessed using global and local MSE and MAE measures. For the construction of training and test sets, four time series (number of: accumulated infected cases, new cases, accumulated deaths and new deaths) from each country are used, which are started on the day of the first confirmed infection case. In order to soften the sudden transitions between samples, a moving average filter with a window size 3 and a normalization by maximum value were used. It is intended to make the model’s predictions available online, collaborating with the fight against the pandemic.

show abstract

“…The units do the operations using only the input data received from the connections. The intelligent behavior of the network comes from the iterations between these units [3][4]. Fig.…”

Section: Introductionmentioning

confidence: 99%