Search

In order to solve the problem of printed Uyghur document image retrieval, a two-level matching method from coarse to fine is proposed. Without word recognition, the character of word image is directly used for fast and accurate document retrieval. Firstly, 100 Uyghur document images are segmented into word sets by using the improved morphological gradient algorithm, and then the contour matching of 8 keyword images is performed by using the contour moment feature to obtain the query result set. Finally, MB-LBP and other features are extracted from the word images in the set. The trained OCSVM classifier is used to classify the feature data, and the result is returned to the user. The experimental results of 8 keyword images in 100 document images show that the average accuracy is 64% and the average recall rate is 49%. The experimental results show that efficiency of the keyword-based retrieval method for Uyghur document image.

In the modern era of rapid development of information technology and mobile Internet, mobile terminals represented by Android smart phones are applied to various fields of life, and the number of users is far more than that of traditional PC terminals. Mobile financial applications marked by mobile banking and mobile payment have higher access control requirements, while traditional text passwords, dynamic passwords and verification codes have hidden dangers of being stolen and copied. How to apply biometrics to mobile identity authentication with high security requirements has become a hot issue for scholars all over the world. Lip reading and speech biometrics are studied in this paper, and a multi-biometric fusion authentication system based on both on Android smart phone platform is developed.

Recent years, some applications of single biometrics recognition have been developed in our lab. This paper summarizes some theories and algorithms involved in those development processes. First, we give some comparison between those biometrics recognition technologies. Then biometrics database we used in the research work is introduced. After this, mainly introduces some common biometrics recognition processes, which are basically include biometrics signal acquisition, preprocessing, feature extraction, biometrics pattern matching and recognition results. Finally, give a summery about this biometric recognition.

With the security requirements of information age, multimodal biometrics has been becoming the most potential technologies for identity recognition. This paper demonstrates the inevitability that multimodal biometrics replace single mode by performance comparison firstly. Then, multimodal biometrics, acquisition methods and the common sensors are listed, different fusion levels are analyzed, fusion methods are given by reference in levels. Additionally, multimodal biometrics systems in application are classified and introduced by platform, meanwhile application scenarios are showed. Finally, we conclude that multimodal biometrics technology must be the research trends, and make prospect for existing deficiencies in its research, especially template protection related to the reliability and safety of systems.

As a dynamic feature, lip-reading is not easy to imitate, and it can collect face information and voice information at the same time, so it has high convenience. The individuality of lip-reading can be used as the basis of speaker's identification. We can also integrate lip-reading and face information, lip-reading and voice information, which can greatly enhance the anti-counterfeiting ability of the recognition system while improving the accuracy of the recognition system. This paper mainly introduces the application of lip-reading technology in speaker recognition, introduces in detail the content and method of lip-reading technology, and the characteristics of lip-reading technology in speaker recognition, aiming to arouse people's attention and interest in this research direction and promote the progress of relevant issues..

Since the multispectral palmprint has more spatial and frequency characteristics than the natural light condition, we can extract more feature information and obtain higher recognition accuracy. In recent years, many scholars have committed themselves to multispectral palmprint recognition studying. In this paper, a novel robust p-norm regularization model is proposed to implement the multispectral palmprint recognition. Then, a proximal iterative reweighted algorithm is employed to solve this model. Finally, we utilize the weighted fusion strategy to calculate the representation residual and carry out the recognition task efficiently. Extensive experiments on PolyU multispectral palmprint database illustrate that the proposed algorithm can achieve outstanding recognition accuracy and outperform some conventional representation methods.

The algorithm of classification is one of the important problems to be solved in the field of chip manufacturing, which has a great impact on the efficiency of process such as subsequent chip packaging. According to the requirement of intelligent control of chip production system, chip classification algorithm based on extreme learning machine (ELM) is studied. In this paper, we use image edge gradient information as feature vector and use ELM to classify the chip. In order to improve the speed of the algorithm, we use image pyramid to down-sample the image first. The final experimental results show that, in small-scale testing, our algorithm can achieve 100% accuracy and it is insensitive to illumination changes. When the image rotates, our method can achieve more than 93.3% accuracy.

Palmprint authentication technology can solve the information security problem of mobile devices effectively. Based on smartphone, a new palmprint acquisition scheme and an improved authentication algorithm are proposed in this article. Key points and ROI region of palmprint are drawn in the smartphone interface to guide acquisition. In order to overcome the high computational complexity of traditional algorithm, ROI image get in acquisition is down-scaled, then the feature vectors extraction of ROI can be down-sampled with Log-Gabor filter and encoded with a new coding method. Simulation results show that our acquisition scheme can improve the quality of palmprint sample effectively, meanwhile the improved authentication algorithm has achieved good accuracy and timeliness.

In the past few decades, biometrics have developed rapidly, and such achievements are inseparable from the support of a variety of single-mode and multi-modal biometric databases. The use of a biometric database makes it possible to compare the performance consistency of different recognition technologies. The collection of these databases is a time-consuming and resource-intensive task for researchers, especially for multimodal databases involving multiple biological characteristics. This paper reviews some single-mode and multi-mode biometric databases which are widely used in the world at present.