基于CNN的二维四通道不可分小波滤波器识别

二维四通道不可分小波在图像融合中得到了成功的应用,但其滤波器组的选取都是依靠经验手动完成的,这样会导致图像融合的结果难以达到最优。针对在图像融合的过程中选择融合效果最优的滤波器组的问题,提出基于CNN的二维四通道不可分小波滤波器组的识别方法。该方法利用卷积神经网络强大的学习功能,对构造的大量二维四通道不可分小波滤波器组进行训练。让训练好的网络对滤波器组进行识别,让这些识别的不同类别滤波器组分别作图像融合,并对比其融合效果。实验结果表明,该方法在测试集上的识别准确率达到0.999 5,对训练测试集外的滤波器组同样识别准确,解决了图像融合过程中二维不可分小波滤波器组选取困难的问题。

A Network Model on the Processing of Sound Wave

On the base of auditory neural system, the network model on the processing of the sound wave is presented. The mathematic equation of the network is also discussed. In the network model, in addition to the negative feedback of the neural cell in the output layer, the cell in the input layer excites the corresponding cell in the ontput layer meanwhile it inhibits the lateral cells. The network has its advantage on the processing of sound wave. In addition to filter the noise, it can search the significance frequency segments （Barks）. The ＂channel suppresser＂ feature, the special phenomena of the human ear, is explained based on the model. The learning algorithm of the network model is discussed, too. In the end, an example is introduced about the application of the network.

Adaptive bands filter bank optimized by genetic algorithm for robust speech recognition system

Perceptual auditory filter banks such as Bark-scale filter bank are widely used as front-end processing in speech recognition systems.However,the problem of the design of optimized filter banks that provide higher accuracy in recognition tasks is still open.Owing to spectral analysis in feature extraction,an adaptive bands filter bank (ABFB) is presented.The design adopts flexible bandwidths and center frequencies for the frequency responses of the filters and utilizes genetic algorithm (GA) to optimize the design parameters.The optimization process is realized by combining the front-end filter bank with the back-end recognition network in the performance evaluation loop.The deployment of ABFB together with zero-crossing peak amplitude (ZCPA) feature as a front process for radial basis function (RBF) system shows significant improvement in robustness compared with the Bark-scale filter bank.In ABFB,several sub-bands are still more concentrated toward lower frequency but their exact locations are determined by the performance rather than the perceptual criteria.For the ease of optimization,only symmetrical bands are considered here,which still provide satisfactory results.

Recognition of human face based on improved multi-sample

In order to solve the problem caused by variation illumination in human face recognition,we bring forward a face recognition algorithm based on the improved multi-sample. In this algorithm,the face image is processed with Retinex theory,meanwhile,the Gabor filter is adopted to perform the feature extraction. The experimental results show that the application of Retinex theory improves the recognition accuracy,and makes the algorithm more robust to the variation illumination. The Gabor filter is more effective and accurate for extracting more useable facial local features. It is proved that the proposed algorithm has good recognition accuracy and it is stable under variation illumination.

基于空间转换网络的端到端车牌检测与识别

在复杂场景中,许多现有的车牌检测和识别方面的研究方法存在数据集单一且有限、算法复杂等问题。因此提出了一个端到端的统一网络:残差-空间变换-连接时序分类融合的车辆号牌检测识别网络(LPDR-RSCNet)。该网络结合残差神经网络、空间变压器网络和连接主义者时间分类,联合训练检测和识别模块,以减少中间错误积累。通过在残差神经网络提取特征过程中引入空间变换网络,使特征提取器具有平移不变性、旋转不变性和缩放不变性;在分类器引入连接时序分类,可以自动识别图片标签和特征之间的关系。同时,还可以适应可变长度序列的识别。在中国城市停车场数据集(CCPD)上进行了比较实验,CCPD是一个大规模、多样的中文车牌数据集。实验证明LPDR-RSCNet模型在实际应用中可实现98.8%的识别精度和34 fps的速度,并且相较于YOLO9000、Faster-RCNN、SSD300,具有更好的检测准确度,可满足智能交通系统中对移动车辆实时车牌检测和识别的要求。