Digital Signal Processing
Data Collection for Digital Signal Processing
Modern signal processing applications involve applying tools from signal processing to all kinds of data– from twitter streams to heart rate time series to large image databases. Digital Signal Processing or DSP is a core undergraduate signal processing course in every curricula across the country. Some of the main topics include sampling, quantization, and filtering and processing of discrete-time signals. Many very interesting and relatively modern applications of DSP, like image processing and error-correction coding, are not typically included because they are not part of the classic textbooks. The tools for even more contemporary applications, especially those in “big data,” such as environmental sensing and health monitoring, are completely left out of the standard DSP course.
To address this, we are incorporating a data collection project into Digital Signal Processing. In teams, the students create a data collection plan for their chosen sensor type, including self-activity monitors, temperature sensors, cameras, and audio recorders. Using these data they have actively collected, they will then explore various contemporary signal processing tools to analyze and draw conclusions from the data. Relating the tools to their personal experience is a powerful way to make the material immediately accessible and relevant, which will simultaneously make it easier for them to learn and recall the material, as well as help them to think critically about the models, approaches, and applications. This results in the students developing critical thinking skills with regards to the application of well-known algorithms to real data.
Reports from Fall 2013 and 2014 can be seen here:
The student projects ranged from replicating reverb of different places around campus using audio sensors and impulse response theory; to a connect four AI system that can play in real time with image processing; to a systems analysis of our body measuring the input of power on a bicycle and the output of heartrate; to comparing Fitbit to Android sensors in step counting; to hand gesture recognition, an important capability for modern technologies such as Google Glass. The students learned advanced techniques to tackle each of their own chosen problems, and through this they had a broad exposure to a variety of tools for modern signal processing.
Reactions from students:
“[The project] boosted my confidence in being able to tackle problems quickly. The freedom [given in the project] to try something new was a very exciting and rewarding experience.”
“The project was much more rewarding than homework/exams since it allowed us to work on an application of DSP that we found interesting. I really enjoyed the design aspect and having a finished product that we can discuss with other students, professors, employers, etc. Implementing a design was by far the best way to thoroughly understand some of the concepts learned in class.”
“The best part about doing a project for class was the sense of ownership of my education that it provided. It was a great “hands on” form of learning that we were able to govern ourselves, so we were able to chose something that we would find most valuable.”
Laura Balzano, Assistant Professor, College of Engineering