In Digital Signal Processing, students experiment with sensors for creative projects
In Prof. Laura Balzano’s Digital Signal Processing and Analysis course (EECS 351), students created projects with a variety of goals, from improving a non-invasive method of assessing traumatic brain injury patients to classifying spoken language to autonomously evaluating a musician’s piano playing. The Winter 2022 course featured an enrollment of approximately 70 students.
“The goal of EECS 351 is to learn the fundamentals of Digital Signal Processing and understand the myriad ways we represent and process signals and data on a computer,” Balzano said.
In the course, students to gain firsthand experience with sensors and many signal processing techniques. They design and carry out a data collection and analysis project on a topic of their choosing, which gives them experience on how algorithms work in real life on real data.
“The creativity and range of the projects was outstanding, as usual,” Balzano said.
The creativity and range of the projects was outstanding, as usual.
Prof. Laura Balzano
The course incorporates image processing and machine learning techniques and is good preparation for students looking to take additional 400-level courses that build off the material, such as EECS 452 (Digital Signal Processing Design Laboratory), 453 (Principles of Machine Learning), 455 (Wireless Communications Systems), and 460 (Control Systems Analysis and Design).
Check out the summaries of all the projects below:
By Faulker Bodbyl-Mast, Jackson Carroll, and Kelman Wolfostin
Goal: Create an audio plugin that can isolate a source sound (i.e. human speech) from the acoustical sound, or reverberation, that voice excites in a room. This has many relevant applications across the audio industry, particularly for those working for the film and music industries.
By Aditya Bhatnagar, Shahad Alsayyad, Sergio Goodwin, Ali Robinson, and Yiwei Zeng
Goal: Create a program that can operate lights in a house based on hand cues, similar to a smart home system. The number of fingers a user holds up to a live camera feed identifies the target light, and a thumbs up or thumbs down gesture changes the state of the light.
AMP Simulation Project
By Logan Kibler, Sam Smith, and Hunter Adams
Goal: Create an accurate recreation of the sound you hear running a guitar signal through a AB763 Fender Twin Reverb guitar amplifier. Guitar amps such as this one can weigh up to 100 pounds and can be tedious to move around. Also, audio engineers and producers may want to try out the sound of an amplifier that they do not actually own, and a virtual recreation would be a cost-effective and quick way to do so.
Detecting Noise in Eye-Impedance During Traumatic Brain
By Parker Stogdill, Sean O’Connell, Wes Cummings, and Eason Chang
Goal: Detect and remove impulse noise for improved non-invasive evaluation of traumatic brain injury patients. The condition of traumatic brain injury patients is traditionally determined by drilling through the skull and inserting a sensor into the brain to measure the blood pressure of the brain. As an alternative the Tiba Lab at Michigan Medicine is doing research into using eye-impedance, a non-invasive measure, to track the patient’s condition. This team is working to improve the success of this non-invasive procedure.
Computer-Aided Melanoma Detection
By Kabir Deol, Dominic Dadabbo, Ryan Downey, and Joseph Wendt
Goal: Accurately detect melanoma on a person’s skin through a computer-aided algorithm people can use in their own home. This could help save costly visits to the dermatologist while helping improve early detection of melanomas. In addition, machine learning algorithms have historically excluded people of color when training their datasets. This team trained their data to detect melanoma on a variety of skin pigmentations, ensuring no groups would be left out.
Handwriting to LaTex
By Emilia Psacharopoulos, Enakshi Deb, Maura Mulligan, James Wishart, and Ritika Pansare
Goal: Translate written characters from a whiteboard into a well-formatted LaTeX document. Many lectures are recorded for students who are unable to attend in person. However, it can be difficult to read writing on whiteboards via the recording. By translating the written content automatically, this program improves accessibility for all.
MP3 to MIDI Converter
By Cody Dempster, Wes Mackey, Kaitlyn Nowak, Laurel Saxe, and David Suh
Goal: Take any MP3 audio file and convert it to the MIDI file format. The MIDI file format can then be used to obtain the sheet music for the song. The converter would be most helpful for beginner musicians who would like to learn to play a song that they heard.
Classifying Dogs vs Cats from Images
By Iman El-Bawab, Ankita Maahajan, and Shwera Pati
Goal: Create a highly accurate model that can predict whether an image is a cat or a dog. The model can then be expanded to more complex systems, such as identifying emotion based on facial expressions in pictures, or distinguishing between pedestrians/hazardous objects in the road scene for autonomous vehicles.
Noise Cancellation Project
By Sandilya Sai Garimella, Andrew Elliott, and Jocelyn Gu
Goal: Use active noise reduction to remove loud, unwanted noise, so as to create less noisy conditions for work, study, and daily life. The active noise reduction function is to generate reverse sound waves equal to the external noise through the noise reduction system, neutralize the noise, and achieve the effect of noise reduction.
By Servando Garza, Zion Studivant, Edward Ted Ivanac, Marco Tulio Giachero Pajaro, and Matthew Perez
Goal: Understand the different methods in which images can undergo audio-based sonification and create a musical application that can be made of the time-varying data from the images. Image sonification is the process of converting non-time varying data in images into a time-varying form.
Rating Practice Piano Pieces
By Stefanos A Frilingos, Blake A Hall, Wentao Xu, Xiangdong Wei, and Haoliang Cheng
Goal: Adopt multiple features of the music signal processing models to automate the evaluation of a large set of piano pieces. Autonomous rating of music performance could better the performers’ understanding of playing techniques and provide a flexible tool for the evaluation of autonomous music generation
By Aashish Karikrishnan, Daniel Li, Andrew Lyandar, and Shantanu Purandare
Goal: Classify spoken languages by using a multi-language identification system and audio recordings from at least three languages. In addition to classifying distinct languages, the model is used to sort each language and dialect into categories.
By Jackson Muller, Andy Zhang, Dawson Hartman, and Justin Yu
Goal: Assisting people in solving Wordle. In Wordle, players have six tries to guess a randomly chosen five letter word. The team created a program that helps people evaluate the quality of their starter word. They also created a User Interface that assists the player by suggesting words based on the player’s previous entry.
Using DSP to remaster old songs
By Richard Shen, Michael Sun, and Hyugo Weicht
Goal: Digitally remaster songs by employing filtering techniques to remove noise and amplify the original song. The team also implemented a general equalizer where a user can adjust the “low’s, mid’s, and high’s” of the filtered recording manually. This will improve the preservation of old recordings and ensure the artistry of past musicians can continue to be enjoyed.
Classification of Underwater Mammals
By Izzi Nolan, Jason Ribbentrap, and Sebastian Sulborski
Goal: Identify dolphins and whales based on their underwater songs. The team created filters to minimize noise from external factors to improve the accuracy of identification. This program can be used to alert motorized boats to the presence of dolphins and whales.
By Qianxu Li, Tianwei Liu, and Raj Patel
Goal: Create a system that can effectively isolate the audio of individual speakers and remove noise accordingly. This system can be used in various applications such as auto caption, noise cancellation for zoom, real-time audio selection, and more.
Low-Resolution Image Processing with a Quadcopter
By David Li, David Engel, and Hengrui Tian
Goal: Test image processing of unmanned aerial vehicles (UAV) through red color detection, circle detection, and chessboard coordinates determination. Image processing enables UAV’s to make decisions in real time based on the environment. For example, an UAV can use its camera to detect objects, which makes drone delivery a possibility.
By Jedidah Pienkny, David Pulido, and Erik Radakovich
Goal: Create a stem player, which takes digital music files and manipulates them in order to isolate the various parts of their makeup: drums, bass, vocals, and more. This is useful for music enthusiasts who like remixing their favorite songs, or for testing sound systems to determine optimal or sub-optimal performance in specific ranges.