Wyniki wyszukiwania - BazTech

Ograniczanie wyników

Znaleziono wyników: 2

Liczba wyników na stronie

Wyniki wyszukiwania

Wyszukiwano:
w słowach kluczowych: temporal continuity

Sortuj według:

Ogranicz wyniki do:

Speech Enhancement Based on Constrained Low-rank Sparse Matrix Decomposition Integrated with Temporal Continuity Regularisation

Sun Chengli, Yuan Conglin

Archives of Acoustics

2019

Vol. 44, No. 4

681--692

Speech enhancement in strong noise condition is a challenging problem. Low-rank and sparse matrix decomposition (LSMD) theory has been applied to speech enhancement recently and good performance was obtained. Existing LSMD algorithms consider each frame as an individual observation. However, real-world speeches usually have a temporal structure, and their acoustic characteristics vary slowly as a function of time. In this paper, we propose a temporal continuity constrained low-rank sparse matrix decomposition (TCCLSMD) based speech enhancement method. In this method, speech separation is formulated as a TCCLSMD problem and temporal continuity constraints are imposed in the LSMD process. We develop an alternative optimisation algorithm for noisy spectrogram decomposition. By means of TCCLSMD, the recovery speech spectrogram is more consistent with the structure of the clean speech spectrogram, and it can lead to more stable and reasonable results than the existing LSMD algorithm. Experiments with various types of noises show the proposed algorithm can achieve a better performance than traditional speech enhancement algorithms, in terms of yielding less residual noise and lower speech distortion.

Load balancing algorithm for parallel vision system using cots pcs and networks

Kim Y., Kim B.

Image Processing & Communications

2002

Vol. 8, no 2

43-56

We developed a real-time parallel vision system (PAVIS) along with suitable parallel vision algorithms. PAVIS is designed for cost-effectiveness and flexibility, which is composed of several isotypic commercially off-the-shelf (COTS) PCs with an Ethernet switch that constitute a logically fully connected parallel system. Image partitioning method using the equal area criterion has been used for load distribution in parallel vision processing. Since the load is not the pixels but the feature points for high-level vision operations, we propose an image partitioning method named perpendicular image partitioning (PIP), which partitions image data not by the area of image but by the number of features. PIP functions as an efficient load distribution method, since image partitioning and load balancing are simultaneously performed. Real-time vision algorithm for corner detection and feature matching are also devised. Since the objects on the temporally successive sequence of image have motion continuity, we can speedup vision operations by reducing search area for corners or matching pairs along the motion trajectories found from the prey ions image frames. PAVIS is applied to the real-time depth determination problem for mobile robot navigation. The experimental results confirm the validity of proposed algorithms, and its real-time performance within the frame rate.