A5060936410- Video Analysis and Summarization
- Advanced Image Processing Techniques
- Generative Adversarial Networks and Image Synthesis
- Speech and Audio Processing
- Robotic Path Planning Algorithms
- Music and Audio Processing
- Vehicle License Plate Recognition
- Animal Vocal Communication and Behavior
- Handwritten Text Recognition Techniques
- Advanced Manufacturing and Logistics Optimization
- Indoor and Outdoor Localization Technologies
- Advanced Adaptive Filtering Techniques
- Computational Geometry and Mesh Generation
Tsinghua University
2019-2021
Institute of Microelectronics
2019-2021
The polyphonic sound event detection and localization (SELD) system indicates the temporal onset offset time of events to be detected tracks spatial location acoustic source. It involves two processes, (SED) estimation direction arrival (DOA). However, previous models only extract features by simply stacking convolutional layers, thus leading problems, one is that network difficult deepen expressive capability model limited, another problem they utilize highlevel features, lack a feature...
This paper proposes a novel model for video generation and especially makes the attempt to deal with problem of from text descriptions, i.e., synthesizing realistic videos conditioned on given texts. Existing methods cannot be easily adapted handle this task well, due frame discontinuity issue their text-free schemes. To address these problems, we propose recurrent deconvolutional generative adversarial network (RD-GAN), which includes (RDN) as generator 3D convolutional neural (3D-CNN)...
It is known that the signal noisy and susceptible to multipath interference in indoor positioning, resulting a significant error processing of signal. The RSS- assisted cross-correlation (RACC) method can reduce noise eliminate certain extent, but too environmentally sensitive. Therefore, this paper, an effective way using deep neural network proposed address problem. Accordingly, performance AutoEncoder reduction elimination are discussed. To achieve better results, four models put forward,...
Optimization of the path planning algorithms is an important direction in many research fields, such as robot and map. In this paper, a noval graph lossless compression method presented for layout partitioning to shorten time planning. This algorithm designed files, based on polygons each layer layout. It can be loaded various search optimization methods find bypassing between two points cutting line. The proposed not only improves partition efficiency significantly, but also makes it easier...