Intonation recognition for indonesian speech based on fujisaki model

Kurniawan

Dwiantoro

et al. 2022

IJ-AI

Self Cite

The oxygen content of the gas-fired boiler flue gas is used to monitor boiler combustion efficiency. Conventionally, this oxygen content is measured using an oxygen content sensor. However, because it operates in extreme conditions, this oxygen sensor tends to have the disadvantage of high maintenance costs. In addition, the absence of other sensors as an element of redundancy and when there is damage to the sensor causes manual handling by workers. It is dangerous for these workers, considering environmental conditions with high-risk hazards. We propose an artificial neural network (ANN) and random forest-based soft sensor to predict the oxygen content to overcome the problems. The prediction is made by utilizing measured data on the power plant’s boiler, consisting of 19 process variables from a distributed control system. The research has proved that the proposed soft sensor successfully predicts the oxygen content. Research using random forest shows better performance results than ANN. The random forest prediction errors are mean absolute error (MAE) of 0.0486, mean squared error (MSE) of 0.0052, root-mean-square error (RMSE) of 0.0718, and Std Error of 0.0719. While the errors using ANN are MAE of 0.0715, MSE of 0.0087, RMSE of 0.0935, and Std Error of 0.0935.

Section: Introductionmentioning

confidence: 99%

The prediction of the oxygen content of the flue gas in a gas-fired boiler system using neural networks and random forest

Kurniawan

Dwiantoro

et al. 2022

IJ-AI

Self Cite

“…Machine learning has been utilized to recognize patterns [11]- [13], estimate and control the output of a process [14]- [16] and predict industrial process variables [17]- [20]. One type of machine learning method is the radial basis function network (RBFN) [21].…”

Section: Introductionmentioning

confidence: 99%

Radial basis network estimator of oxygen content in the flue gas of debutanizer reboiler

Sembodo

Dwiantoro

et al. 2022

IJECE

Self Cite

The energy efficiency in the debutanizer reboiler combustion can be monitored from the oxygen content of the flue gas of the reboiler. The measurement of the oxygen content can be conducted in situ using an oxygen sensor. However, soot that may appear around the sensor due to the combustion process in the debutanizer reboiler can obstruct the sensor’s function. In-situ redundancy sensors’ unavailability is a significant problem when the sensor is damaged, so measures must be made directly by workers using portable devices. On the other hand, worker safety is a primary concern when working in high-risk work areas. In this paper, we propose a software-based measurement or soft sensor to overcome the problems. The radial basis function network model makes soft sensors adapt to data updates because of their advantage as a universal approximator. The estimation of oxygen content with a soft sensor has been successfully carried out. The soft sensor generates an estimated mean square error of 0.216% with a standard deviation of 0.0242%. Stochastics gradient descent algorithm with momentum acceleration and dimension reduction using principal component analysis successfully improves the soft sensors’ performance.

Automatic recognition of Indonesian declarative questions and statements using polynomial coefficients of the pitch contours

Sairyo

Furui

et al. 2009

Acoust. Sci. & Tech.

Self Cite

Abstract:We propose an automatic utterance type recognizer that distinguishes declarative questions from statements in Indonesian speech. Since utterances in these two types have the same words with the same order and differ only in their intonations, their classification requires not only a speech recognizer, but also an intonation recognizer. In this paper, the most important utterance part for distinguishing those two types is first identified by perceptual experiments. Then, an utterance type recognizer using that part is proposed, where polynomial expansion is used as a feature extractor and a neural network is used as a classifier. We evaluated our method using Indonesian speech database including 29 pairs of sentences of those two types, each of which uttered by 35 speakers. It was proved that final word and final-two-syllables are equally effective for the discrimination of each utterance. The proposed recognizer achieved the best accuracy of 89.1% when the order of polynomial expansion was three and the neural network was a linear perceptron.