Pu Yuanyuan

Feature Extraction and Analysis for Scientific Understanding of Visual Art

The authentication and classification of visual arts used to be depended on the judgment of the critic according to his experiences and knowledge about the artwork and artist. Recently, with the development of computer and image processing technology, the processing of digital visual artworks has become possible. More and more museums and libraries have their collections to be digitalized, which makes the study of problems haunted the art historians based on these digital images available. In this paper, the research of visual arts based on the computer and information of paintings, which can be called scientific understanding of visual arts, has been brought out. Some features, including multi-scale amplitude, non-stationarity of artworks, anisotropy of artworks and correlation of coefficients of the Curve let transform among scales are extracted based on the Curve let transform to evaluate the styles of different artists and some other respects of artworks. The relations between the styles of visual arts and these features are also stated, and the similarities of these styles are also qualified by comparing these features. It is apparent that each feature reflects different characteristics of the painting technique.

Speaker Recognition Based on Yunnan Minority Accent Mandarin Speech Corpus

Based on the fact that most Yunnan minority people speak mandarin with their native accents, a linguistic minorities accents mandarin speech corpus for non-native speech signal processing is constructed in this paper. Based on this corpus, a Bayesian network for speaker recognition is designed and the method of how to setup the parameters is introduced too. According to the task of speaker recognition with text independent, the experiments are carried on the Yunnan minority accent Mandarin speech corpus and compare the results of the different Bayesian networks parameters, and different the dimension of MFCC.

A Naxi speech synthesis system based on pitch-synchronous overlap-add

Naxi is a tone language spoken by the Naxi nationality of China. Based on pitch-synchronous overlap-add technology, a Naxi speech synthesis system which can work in real time on a PC is developed according to the acoustical features of Naxi. In this system, monosyllables and four tones are selected as the basic speech unit. It reads in Naxi Pinyin text and converts it to fluent speech. Furthermore, a prosodic model including tone, duration and pause rules is constructed and applied in the Naxi speech synthesis system successfully.