11-755 MLSP, Projects page

11-755 MLSP -- PROJECTS

Team members	Project Ideas
Joseph Fernandez, Gopala Anumanchipalli, Qi Mao, Xiaonhan (Libra) Huang, Ying-Chang (Allen) Cheng	Statistical Klatt Parametric Synthesis
Ryan Walsh, Andrew Lam, Han Joo Chae, Steven Thompson	Augmented Reality / Seam Carving / Audio Fingerprinting
David Leong, Brendan Jou, Aneeb Qureshi, Jessica Liao	Content-aware resizing for video applications
Tianhua Lu, Xun Jiang, Bingran Liu, Peikang (Alpha) Song	Voice transformation with Canonical Correlation Analysis
Xu Wu, Hao Jin, Juefei Xu, Shih-Hao Wen	Talking Karaoke
Chanwoo Kim, Sang Hyun Lee, Geoffrey Dixon-Ernst	Sound source separation and missing feature enhancment
Udhaykumar Nallasamy, Nikoloz Tsimakuridze, Fred Fishbeck	Voice transformation
Siddesh Mehra, Krishnan Ramkumar	Image segmentation
Mario Berges	Non-intrusive load monitoring
Ashwati Krishnan	Counting blood cells in Cerebrospinal Fluid
Siddharth Hazra	Determining Music Tablature
Michael Chuang	Finding right camera angles in sports video
Rachel Cawley, Tina Milo	Image Deblurring
Chih Chao Chen, Jinyin Zhang, Qingxing Zhang	Voice Transformation
Cheng Chih Yen	Face detection

Project presentation schedule

Some possible projects

Sound classification: Can you distinguish between different types of cars by the sound they make? (Pick your collection of sounds to distinguish)
Image deblurring: Have you ever taken a photograph, only to have it all blurry because your hand shook? Wouldn't it be great if we could unblur the image?
Talking Karaoke: Can you replace the singer's voice in a song with your own? While never actually singing?
Diarization: You have a recording of the audio in an episode of “Friends” (for example). Can you automatically find all the regions where each of the characters speaks? Can you automatically segment out the jokes? Can you do the same from only the video? What about if we had both?
Gait Recognition: From Doppler ultrasound recordings of a person walking, can you determine who it is?
Image Modification: Patch transforms are a new technique for modifying images, that involve both signal processing and some complex machine learning techniques. An implementation of patch transforms would be a very impressive project.
Speech from Ultrasound: The user mimes speech. We generate actual speech from this. This has been a long-term desire of many companies and espionage agencies..
3D from 2D: We have many photos of a person's face. Can we compose a 3D model of the face from these photos?
3D from 2D animation: Find the closest match to a given 2-D stick animation in a collection of video. Use this to generate 3-D animation (Eakta Jain's project).

Fernando De La Torre also has some fun projects.