Selected Projects

 I designed a software for analysing keywords and plotting word-cloud from users' posts in Renren. It was written in Python.

Main work :
  • Crawling user's posts
  • Parsing and analysing keywords
  • Plotting word cloud images based on integral image algorithm
  • experimenting on a dataset of 10 million sina micro-blogs with my parser 
Toolkits include:
  • python-packages: urllib, requests (crawling)
  • python packages: PIL, matplotlib (plotting), scipy, numpy (calculating)
  • NLPIR, a popular Chinese NLP toolkit
It was my first project as an undergraduate. Renren was the most popular social network among college students at that time — just a Chinese version of Facebook.  But most of us turned to 'Moments' provided by Wechat and Renren is now out of date. How time flies!

Image Classifier based on LDA

   the 27th Participation in Research Program in SJTU (Supervisor: Jie Guo)

I developed an image classifier in Visual Studio (C++).

 It contained following parts: 
  • Extrating SIFT features of images
  • Clustering features by k-means algorithm and forming a 'dictionary'
  • Calculating the counts of features for each image
  • Classifying with latent Dirichlet allocation (LDA)
I experimented on five kinds of images in Caltech101 dataset and achieved satisfying results.


Crack Visual CAPTCHA

(course project for Introduction to Artificial Intelligence

 I led a group of three and implemented state-of-art algorithms for      cracking visual CAPTCHA.  It was developed with MATLAB.

Main work: 
  • crawled images from SJTU JAccount logon page
  • image preprocessing: segmentation, de-rotation, size       normalisation, etc.
  • image labelling (manually)
  • experimented with two classification methods 
    • k-Nearest-Neighbour algorithm
    • feedforward neural network 
    • for 10-fold cross validation, average error rate:  4.5%

Network Sniffer Software

 (course project for Computer Communication and Network 

I independently designed a network sniffer software with pretty UI.
It was written in Visual Studio (C#).

Main functions include: 
  • Listening to a network interface card
  • Snifferring and Analysing all kinds of network packets, including  TCP, UDP, ARP, etc. 
  • Filtering packets by protocols/ MAC/ ports/ IP address

Magic Beats : an iOS App

 (course project for Software Engineering)
I led a team of four and we designed and simulated an Application named "Magic Beats" on iOS platform. It was written in Objective-C.

The application provided several functions including:
  • Recording, Saving and Sharing to other platforms
  • Changing voice, either manually or with preset modes (Girls/ Men/ Robots)
  • Registering and Managing user accounts


Detection against Shrew DDoS Attack based on DSP methods

(course project for Digital Signal Processing  

I led a team of three and developed a set of detection methods against Shrew DDoS Attack. 

My work included:
  • Simulating normal network traffic with NS2, and Generating periodic attack flow
  • Choosing appropriate filters (like Kaiser Window)
  • Analysing frequency spectrum and energy allocation
  • Designing detection method and setting thresholds