Search

Your search keyword '"Yang, Junan"' showing total 100 results
Start Over Author "Yang, Junan"
100 results on '"Yang, Junan"'

10. Improved features using convolution-augmented transformers for keyword spotting

Author

Wang Yi, Yang Junan, Liu Jingtao, Chen Qiang, and Li Song

Subjects
keyword spotting, attention, convolutional neural networks, transformers, Information technology, T58.5-58.64
Abstract

Transformer can effectively model long rang dependency, but suffer from uncapable to extract local feature patterns. While CNNs exploit local features effectively. In this paper, we seek to combine convolution and Transformers improves over using them individually, and propose improved features using convolution-augmented transformers for keyword spotting. The convolution-augmented transformers are constructed with a ResNet front-end and a convolution-augmented transformers back-end in series. Using this improved feature for keyword spotting task. The results show that the improved features using convolution- augmented transformers can yield at least 3% improvement compared with other features.

Published
2022
Full Text
View/download PDF

44. 压缩感知非凸优化超宽带信道估计方法研究

Author

Wang Shafei, Wen Zhijin, and Yang Junan

Subjects
Mathematical optimization, General Computer Science, Spectral density, Reconstruction algorithm, Compressed sensing, Norm (mathematics), Convex optimization, Band width, Convex function, Non convex optimization, Engineering (miscellaneous), Algorithm, Computer Science::Information Theory, Mathematics
Abstract

Considering the complicated transfer propagation and low power spectrum density of ultra-wideband (UWB) signal, accurate estimation of UWB channel is crucial. However, it is difficult to sample UWB signal directly for its band width is ultra wide. Compressed sensing (CS) provides a feasible method for low speed sampling. Existing CS-based UWB channel estimation method generally adopts convex optimization form using l 1 -norm restriction, or non-sparse form using l 2 -norm restriction. These methods have a weak restriction on sparseness of objective vector, and l 0 -norm which is the sparsest lacks of effective reconstruction algorithm. To solve these problems, a CS-based UWB channel estimation method based on non-convex optimization is devised in this paper. Firstly, objective function is set as non-convex optimization form using l p -norm restriction; original non-convex function is then combined as convex function form, using the property that convex function obtains extremum easily. Convex function is used to approximate non-convex function at each iteration, for objective vector reconstruction and original channel estimation. Since l p -norm is closer to l 0 -norm, it has stronger restriction on sparseness of objective vector. Experimental results show that proposed method can effectively lower reconstruction error compared to existing CS-based UWB channel estimation method.

Published
2014

46. Is There an Optimal Strategic Oil Reserve for Each Country? A Study Based on the Game Theory

Author

Yang, Junan and Cong, Ronggang

Abstract

In generally, there is a phenomenon of “free rider” in the establishment of national oil reserves for different countries, which means that they have the tendency of underestimating the strategic oil reserves. This paper attempts to explain this phenomenon from the perspective of non-cooperative game theory. It also analyzes the establishment of strategic oil reserve among different countries based on the coalition game theory and presents the core solution for it. The results show that based on a certain constraint mechanism, it is feasible for different countries to establish their own suitable strategic oil reserves in theory and practice.

Published
2014

Catalog

Books, media, physical & digital resources
See catalog results