Search

Sichuan pepper is a traditional and important flavoring of Chinese cuisine. It has attracted increasing interest in recent years owning to its unique taste and aroma. However, some cheap adulterants have been illegally found in Sichuan pepper powder in the market due to merchants trying to cut costs and gain an extra profit. In order to determine the compositions of Sichuan pepper powder quickly and effectively, a direct detection method using near-infrared (NIR) spectroscopy has been developed. 462 samples of adulterated Sichuan pepper powder mixed with different amounts of wheat bran, rice bran, corn flour, and rosin powder were studied. The NIR spectra data was studied using partial least squares (PLS) analysis. The method was found to be capable of predicting the compositions of adulterated Sichuan pepper powder. The determination coefficients of prediction set (Rp2) with the best pretreatments were 0.971 for Sichuan pepper powder, 0.948 for rice bran, 0.969 for wheat bran, 0.967 for corn flour, and 0.994 for rosin powder, respectively. The standard errors of prediction (SEP) were 2.81%, 2.38%, 3.19%, 2.46%, and 1.10%, respectively. The results showed that NIR spectroscopy with chemometrics is a rapid and nondestructive tool for the quantitative analysis of adulterated Sichuan pepper powder.

A full 3D physiological articulatory model was developed to simulate the mechanism of human speech production. In order to control the physiological articulatory model to generate speech sound, we construct a mapping from acoustic objective to muscle activation patterns by simulating the babbling process in speech acquisition. After the babbling phase, the mapping from acoustic objective to muscle activation patterns was built up using a two layer neural network. In order to evaluate the mapping、 muscle activation patterns were estimated from a known speech sound, and then were used to drive the physiological articulatory model to synthesize speech. The comparison results of the acoustic space and articulatory space showed that the proposed mapping-based control strategy was feasible for model control.

Supervisor:党 建武, 情報科学研究科, 博士