Tsinghua Science and Technology最新文献

A CMOS triode transconductor was developed with common mode feedback suitable for operating in low-voltage and low-power applications. The design is based on a body-driven input stage with feedback loops to extend both the signal input range and the tuning capability. The effective transconductance of the body-driven triode stage is increased using a partial positive feedback technique which also partially solves the problem introduced by the small transconductance. This design uses the UMC 0.18 μm CMOS process. Simulations show the transconductor operated with 1 V supply voltage has less than −55 dB total harmonic distortions (THD) in the complete tuning range (0 V ≤ V_cont ≤ 0.43 V) for a 1 MHz 0.8 V_p-p differential input. The power consumption is 70 μW for a 0.43 V control voltage.

Information is an important issue in financial markets since information advantage leads to increased profits. This paper investigates whether those with less risk aversion or more risk aversion would like to be informed when there is private information which can be acquired for a certain cost. The CARA utility function and normal return assumptions were used to get the closed form solution of the absolute risk aversion coefficient which is indifferent to whether a person wants to be informed, which shows that the less risk aversion investor would like to be informed. The results further show that in financial markets, those people with lower risk aversion (for instance, institutional investors) are more favorable to information and price discovery of risky assets, which is consistent with intuition as well as empirical studies.

The sparse representation-based classification algorithm has been used for human face recognition. But an image database was restricted to human frontal faces with only slight illumination and expression changes. Cropping and normalization of the face needs to be done beforehand. This paper uses a sparse representation-based algorithm for generic image classification with some intra-class variations and background clutter. A hierarchical framework based on the sparse representation is developed which flexibly combines different global and local features. Experiments with the hierarchical framework on 25 object categories selected from the Caltech101 dataset show that exploiting the advantage of local features with the hierarchical framework improves the classification performance and that the framework is robust to image occlusions, background clutter, and viewpoint changes.

This paper presents a method for locating text based on a simplified pulse coupled neural network (PCNN). The PCNN generates a firings map in a similar way to the human visual system with non-linear image processing. The PCNN is used to segment the original image into different planes and edges detected using both the PCNN firings map and a phase congruency detector. The different edges are integrated using an automatically adjusted weighting coefficient. Both the simplified PCNN and the phase congruency energy model in the frequency domain imitate the human visual system. This paper shows how to use PCNN by changing the compute space from the spatial domain to the frequency domain for solving the text location problem. The algorithm is a simplified PCNN edge-based (PCNNE) algorithm. Three comparison tests are used to evaluate the algorithm. Tests on large data sets show PCNNE efficiently detects texts with various colors, font sizes, positions, and uneven illumination. This method outperforms several traditional methods both in text detection rate and text detection accuracy.

Mutual information (MI) is a basic concept in information theory. Therefore, estimates of the MI are fundamentally important in most information theory applications. This paper provides a new way of understanding and estimating the MI using the copula function. First, the entropy of the copula, named the copula entropy, is defined as a measure of the dependence uncertainty represented by the copula function and then the MI is shown to be equivalent to the negative copula entropy. With this equivalence, the MI can be estimated by first estimating the empirical copula and then estimating the entropy of the empirical copula. Thus, the MI estimate is an estimation of the entropy, which reduces the complexity and computational requirements. Tests show that the method is more effective than the traditional method.

This paper presents an in-depth analysis of the interference strength and required guardband width between coexistent users for distributed orthogonal frequency division multiple access (OFDMA). In dynamic spectrum access networks, the cross-band interference between spectrally adjacent users is considered harmful with frequency guardbands inserted between spectrum blocks to eliminate the interference. However, the strength of the cross-band interference depends heavily on the user heterogeneity in different OFDM configurations. The cross-band interference due to the three user heterogeneity artifacts of power heterogeneity, sampling rate heterogeneity, and symbol length heterogeneity is investigated to determine the required guardband width. Analytical and simulation results show that the greater user heterogeneity requires larger guardbands with the sampling rate heterogeneity having the greatest effect. These results can be used to assist the design of spectrum allocation strategies.

Cross-modal integration processes are essential for service robots to reliably perceive relevant parts of the partially known unstructured environment. We demonstrate how multimodal integration on different abstraction levels leads to reasonable behavior that would be difficult to achieve with unimodal approaches. Sensing and acting modalities are composed to multimodal robot skills via a fuzzy multisensor fusion approach. Single modalities constitute basic robot skills that can dynamically be composed to appropriate behavior by symbolic planning. Furthermore, multimodal integration is exploited to answer relevant queries about the partially known environment. All these approaches are successfully implemented and tested on our mobile service robot platform TASER.

Literature searches on the Web result in great volumes of query results. A model is presented here to refine the search process using user interests. User interests are analyzed to calculate semantic similarity among the interest terms to refine the query. Traditional general purpose similarity measures may not always fit a domain specific context. This paper presents a similarity method for medical literature searches based on the biomedical literature knowledge source “MEDLINE”, the normalized MEDLINE distance, to more reasonably reflect the relevance between medical terms. This measure gives more accurate user interest descriptions through calculating the similarities of user interest terms to rerank the interest term list. The accurate user interest descriptions can be used for query refinement in keyword searches to give more personalized results for the user. This measure also improves the search results for personalization through controlling the return number of results on each topic of interest.

The efficiency of QoS-aware service composition is important since most service composition problems are known to be NP-hard. With the growing number of web services, service composition is like a decision problem on selecting services or/and execution plans to satisfy the users' end-to-end QoS requirements (e.g. response time, throughput). Composite services with the same functionality may have different execution plans, which may cause different end-to-end QoS. This paper presents a model combining semantic data-links and QoS, which leads to an efficient approach to automatic construction of a composite service with optimal end-to-end QoS. The approach is based on a greedy algorithm to select both services and execution plans for composite services. Empirical and theoretical analyses of the approach show that its time complexity is O(mn²) for a repository with n services and an ontology with m concepts. Moreover, the approach increases linearly in time when using an index to search services in the repository. Tests with a repository with 20 000 services and an ontology with 300 000 concepts show that the algorithm significantly outperforms current existing algorithms in terms of composition efficiency while achieving optimal end-to-end QoS.

With the rapid development of Web 2.0, more and more people are sharing their opinions about online products, so there is much product review data. However, it is difficult to compare products directly using ratings because many ratings are based on different scales or ratings are even missing. This paper addresses the following question: given textual reviews, how can we automatically determine the semantic orientations of reviewers and then rank different items? Due to the absence of ratings in many reviews, it is difficult to collect sufficient rating data for certain specific categories of products (e.g., movies), but it is easier to find rating data in another different but related category (e.g., books). We refer to this problem as transfer rating, and try to train a better ranking model for items in the interested category with the help of rating data from another related category. Specifically, we developed a ranking-oriented method called TRate for determining the semantic orientations and for ranking different items and formulated it in a regularized algorithm for rating knowledge transfer by bridging the two related categories via a shared latent semantic space. Tests on the Epinion dataset verified its effectiveness.