Huakang Li

A Spatial Sound Localization System for Mobile Robots

In this paper, we described a robotic spatial sound localization system using four microphones arranged on the surface of a sphercial robot head. The four microphones, with three on the head center level and one on the top, provide spatial cues for sound elevation as well as azimuth. A weighted cross-correlation method based on the model of the precedence effect is used to calculate the time differences and suppress the influence of echoes. To integrate spatial cues of different microphone pairs, a mapping method from the correlation between different microphone pairs to a 2-D map corresponding to azimuth and elevation of sound sources was proposed. Experiments shows the system can provide the distribution of sound sources in azimuth-elevation space for sound sources, even concurrently in reverberant environments.

Robot Position Identification Using Specially Designed Landmarks

In this paper, we proposed a visual position calibration method using colored rectangle signboards for a mobile robot which is designed to work in a room environment. The rectangle signboards are placed in several known positions. By calculate the vanishing points in the image of the signboard, it is possible to obtain the relative direction and distance between the signboard and the robot by a single image. Experiments of position calibration were conducted. The results showed this method is effective and the errors between the set up positions and the experimental results were about 1.7 degrees and 3 cm when the distance to the signboards was set to 100 cm

Topic Detection from Microblog Based on Text Clustering and Topic Model Analysis

This paper raises a Microblog topic detection method based on text clustering and topic model analysis. It solves the problem that the traditional topic detection method is mainly applicable for traditional media text, which is not very effective in handling sparse Micro blog short texts. In consequence of the structural data of the Microblog, which exists rich inter-textual contextual information such as retweets, comments, user hash tag, embedded link URL, we first put forward a feature weight pre-processing method. We also use a clustering algorithm based on word vectors to enrich the feature information of the data. On this basis, we extend the conventional LDA (Latent Dirichlet allocation) topic model to extract the hot topics in the Micro blog data. Compared with the traditional methods, the method raised in this paper is much more effective in the collected text corpus in Sina Microblog.

Semi-sparse algorithm based on multi-layer optimization for recommendation system

Collaborative filter (CF) is the most successful technique in recommender system, which makes personalize recommendations during online interaction. We propose a new Semi-sparse algorithm based on multi-layer optimization to speed up the basic Pearson Correlation Coefficient of CF. Semi-sparse algorithm spares out over-reduplicate accessing and judgement on selected sparse vector to accelerate the batch of similarity-comparisons in one thread. We propose a reduce-vector in thread-pool to restrict the lock using on critical resources in parallelize implementation. Thread-pool is wrapped with Pthreads on multi-core node to make semi-sparse parallelization more easily. A shared zip file is read to cut down messages with Message Passing Interface package. The performance of proposed semi-sparse with multi-layer framework achieved a brilliant speedup in the evaluation of Netflix, MovieLens and MovieLen1600.

ShmStreaming: A Shared Memory Approach for Improving Hadoop Streaming Performance

The Map-Reduce programming model is now drawing both academic and industrial attentions for processing large data. Hadoop, one of the most popular implementations of the model, has been widely adopted. To support application programs written in languages other than Java, Hadoop introduces a streaming mechanism that allows it to communicate with external programs through pipes. Because of the added overhead associated with pipes and context switches, the performance of Hadoop streaming is significantly worse than native Hadoop jobs. We propose ShmStreaming, a mechanism that takes advantages of shared memory to realize Hadoop streaming for better performance. Specifically, ShmStreaming uses shared memory to implement a lockless FIFO queue that connects Hadoop and external programs. To further reduce the number of context switches, the FIFO queue adopts a batching technique to allow multiple key-value pairs to be processed together. For typical benchmarks of word count, grep and inverted index, experimental results show 20-30% performance improvement comparing to the native Hadoop streaming implementation.

Mobile Robot Position Identification with Specially Designed Landmarks

In this paper, we proposed a visual position calibration method using colored rectangle signboards for a mobile robot which is designed to work in a room environment. The rectangle signboards are placed in several known positions. Using Canny Algorithm to detect the edge of color area to calculate the vanishing points in the image of the signboard, it is possible to obtain the relative direction and distance between the signboard and the robot by a single image. Experiments of position calibration were conducted. The results showed this method is effective and the errors between the set up positions and the experimental results were about 1.5 degrees and 3 cm in an array position.

Robot Position Identification by Actively Localizing Sound Beacons

In this paper, we proposed a robot self position identification method by active sound localization. This method can be used for autonomous security robots working in room environments. A system using a AIBO robot equipped with two microphones and wireless network is constructed and is used for position identification experiments. Arrival time differences to the microphones of robot are used as localization cues. To overcome the ambiguity of front-back confusion, a three-head position measurement method was proposed. The robot position can be identified by the intersection of circles restricted by the azimuth differences to different speaker pairs. By localizing three or four speakers as sound beacons positioned on known locations, the robot can identify its self position with an average error of about 7 cm in a 2.5 times 3.0 m2 working space

Recommendation of More Interests Based on Collaborative Filtering

Collaborative Filtering is one of the most important techniques in recommender systems. Current researches on Collaborative Filtering focus on how to improve the accuracy. However, it is of the same importance to recommend more potential interests to users because many of them have more expectations for recommendation list besides the accuracy. Current recommender systems did not address this problem. This paper focuses on how to help users find more interests in the recommendation list. We propose an sampling-based algorithm Probabilistic Top-N Selection to recommend potential interests for users, and propose two metrics, average predicted rating and category coverage, to assess the quality of the recommendation list. Then we conduct a series of experiments on Movie Lens dataset, experimental results demonstrate that our algorithm can significantly improve user experience through providing them with more potential interests.

Robot navigation and sound based position identification

In this paper, we proposed a robot self position identification method by active sound localization. This method can be used for autonomous security robots working in room environments. A system using a AIBO robot equipped with two microphones and wireless network is constructed and is used for position identification experiments. Arrival time differences to the microphones of robot are used as localization cues. To overcome the ambiguity of front-back confusion, a three-head position measurement method was proposed. The robot position can be identified by the intersection of circles restricted by the azimuth differences to different speaker pairs. By localizing three or four speakers as sound beacons positioned on known locations, the robot can identify its self position with an average error of about 7 cm in a 2.5times3.0 m2 working space. A robot navigation experiment was conducted to demonstrate the effectiveness of the position identification system.

An opinion leader perceptual model based on PageRank algorithm

With more and more social public attention from microblog opinion leaders, they have gradually become the essential role to publicize information and influence public opinion in microblog platform, whose ability in monitoring and guiding public opinions also causes public focus. The thesis first proposes to start evaluation about influence of microblog opinion leaders from perspectives of individual activity, node attention degree, and user node quality and so on based on influence characters owned by microblog opinion leaders and constructs the analytical model of microblog information influence combined with PageRank. It can be found from research results that the method proposed in the thesis can analyze accurately and efficiently the influence of information in propagation route of microblog social communication network.