Li Linsheng

A fast image compression algorithm based on SPIHT

The combination of integer lifting wavelet transform with set partitioning in hierarchical trees algorithm has been widely used in the field of image compression. Based on this algorithm, the paper presents a image coding algorithm with lower memory and higher speed. The algorithm further takes Human Visual System into consideration and modifies SPIHT algorithm is according to the characteristic of weighted wavelet coefficients. Experiment result shows that, under the situation of low bit-rate, the new algorithm can prove the visual effect of reconstructed image at a certain degree. In addition, because the new algorithm has low memory requirement, the coding speed accelerates a lot. It has a very broad appliance future in the occasion of high desire of memory and speed.

A novel adaptive motion detection based on k-means clustering

Detecting a high-quality moving object with good robustness in computer vision system has important significance for follow-up task. Researching on the traditional algorithm, this paper proposes a background reconstruction algorithm based on a modified k-means clustering and the Single Gaussian model which could provide an accurate background image through a sequence of scene images with foreground objects. Then based on the statistical characteristics of the background pixels region detects the moving object. Aiming to the effect of dynamic changes of the environment, this paper proposes a method of robust adaptive motion detection Combined with the principle of Mathematical Morphology and Region-labeling. Experiments prove this method can complete the task of moving object detection in complex environment.

Design of Rail Surface Crack-detecting System Based on Linear CCD Sensor

The rail surface crack-detecting system was designed for reducing railway accident due to rail crack. The system adopts linear charge coupled device (CCD) TCD1208AP as image sensor, uses high-speed flash A/D converter AD7821 to collect CCD output video signals, and uses CPLD perform CCD timing generator, A/D converter timing generator, data storage and other control logic. Then DSP executes the image processing, such as noises elimination, edges detection, image segmentation and edges linking, used the improved classical algorithm and morphology algorithm to judge whether the signals are the crack signals or not, and gives display and alarm with sound and light. The whole hardware structure and the software design are introduced in this paper. By experiment, the detecting precision and effectiveness of the system are good for detecting the rail surface crack.

Non-linear predictor based on ANN in speech coding

In this paper, non-linear predictors based on different artificial neural network (ANN) are compared. General radical basis function (GERBF) neural network, a modified RBF network, is introduced. From the comparison with BP, RNN and RBF, it is obvious that GERBF has priority in the prediction of speech signal. The experiment results show: the speech coding systems based on ANN have better synthesized speech than ITUs G721 and the speech coding system based on GERBF has higher mean segmental SNR but less computation than that of other systems.

Kernel Sparse Representation for Object Tracking

Object tracking is a challenging problem to develop an effective model, which can handle appearance change caused by illumination change, occlusion, and motion blur. In this paper, we propose an online tracking algorithm with kernel sparse representation, local image patches of a target are represented by their sparse codes schemes with an over-complete dictionary, and online classifier is learned to discriminate the target. To improve robustness of the algorithm and the performance of the classifier, kernel function is applied on the sparse representation. In addition to, we propose a simple yet effective method for dictionary update. Experiments on challenging image sequences show that the proposed algorithm performs favorably against several state-of-the-art methods.

A blind channel equalization algorithm based on feed forward neural network

In this paper, a new blind channel identification/equalization algorithm based on multiplayer feed forward neural network is proposed, and the transmission function adapted to the neural network is analyzed. The computer simulations show that the convergence performance of new algorithm is better than traditional feed forward neural network algorithm.