Advances in Engineering Education SPRING 2017 A Mixed Learning Approach to Integrating Digital Signal Processing Laboratory Exercises into a Non-Lab Junior Year DSP Course BENJAMIN D. MCPHERON CHARLES V. THANGARAJ AND CHARLES R. THOMAS Roger Williams University Bristol, RI ABSTRACT Laboratory courses can be difficult to fit into an engineering program at a liberal arts-focused university, which requires students to be exposed to appropriate breadth, as well as sufficient depth in their engineering education. One possible solution to this issue is to integrate laboratory exercises with lecture in a ‘studio’ format, in which students apply lecture concepts directly to in-class assign- ments. Another possible solution is to give students ‘take-home’ laboratory assignments. Both of these methods have shortcomings: the studio format takes away valuable lecture time, and the take-home format provides limited access to the instructor. As such, this work presents a mixed learning method that includes lectures and laboratory work in both the studio and take-home formats, implemented in a junior level signal processing course. Students learn skills during lecture in studio laboratory exercises, and apply these skills to two in-depth take-home projects. Students refine their applied skills during projects, thereby informing a better studio lab experience. In order to assess the student’s developed skills, project results are delivered as research papers formatted to comply with IEEE standards, which are submitted for blind review to several faculty members, as well as their peers. Reviewers employ a prescriptive rubric to rate papers as accept/revise/reject and provide associated comments. To assess the success of this mixed learning method, the overall ratings for the research papers from the first project will be compared to the second project, accounting for project complexity. The chief contri- bution of this work is the presentation of a method for providing laboratory instruction in a mid-year DSP course, demonstrating that this method may be adapted for other courses at similar institutions. Key words: Mixed learning, Studio labs, Take-home projects SPRING 2017 1 ADVANCES IN ENGINEERING EDUCATION A Mixed Learning Approach to Integrating Digital Signal Processing Laboratory Exercises into a Non-Lab Junior Year DSP Course INTRODUCTION Conceptions of learning engineering are dependent on the learning environment, with students identifying lecture as an atmosphere of testing and calculation while the laboratory setting is as- sociated with applying and understanding (Lin, 2009). Lecture is the most common format of educational delivery, but most engineering students also require knowledge that comes only as a result of hands-on experiential learning (Feisel, 2005; Abdulwahed, 2009). As a result, laboratory experiences are vital in undergraduate engineering curricula, especially in sophomore and junior level engineering courses, as they prepare engineering students for internships, research experi- ences, and their senior design project. However, laboratory courses can be difficult to fit into an engineering program at liberal arts- focused university, which requires students to be exposed to appropriate non-engineering breadth, as well as sufficient depth in their engineering education. Auxiliary to this fact, undergraduate-focused universities require faculty members to teach, limiting the availability of instructors for laboratory courses. Owing to this resource limitation, many sophomore and junior level engineering courses in liberal arts-focused universities lack a large number of laboratory experiences. In addition, credit limits and accreditation standards may impose a limit to the number of scheduled laboratory experiences at many universities, making alternatives to the traditional scheduled laboratory period attractive. Two reasonable solutions to this dilemma are to integrate laboratory exercises with lecture in a ‘studio’ format, or to give students ‘take-home’ laboratory experiments or projects. The studio format requires students to apply lecture concepts directly to in-class programming assignments on their personal computers (Whitmal, 2002). Implementing labs in this way allows students to learn engineering skills that practicing engineers are expected to know, with one-on-one access to the instructor. This format has the undesirable side effect of reducing the amount of lecture time available when implemented in a non-lab course. In addition, these types of labs lack time for ap- propriate reflection, and are designed as simply instructional without any research or development components. On the other hand, the take-home format involves assigning lab projects for students to complete on their own outside of class (Jouaneh, 2009). This category of laboratory experience allows time for proper reflection, and helps students transition from instructional laboratories to a development and research opportunity, such as they would experience in the engineering profes- sion (Feisel, 2005). Although this format does not detract from lecture time, there is limited access to the instructor and detracts from the desired laboratory environment. Furthermore, this method requires extensive out-of-class tutorials, whose completion is difficult to police. As there are shortcomings associated with both of these methods, this work presents a new method, which combines the positive aspects of each approach. This mixed learning atmosphere 2 SPRING 2017 ADVANCES IN ENGINEERING EDUCATION A Mixed Learning Approach to Integrating Digital Signal Processing Laboratory Exercises into a Non-Lab Junior Year DSP Course includes lectures and laboratory work in both the studio and take-home formats. Students were given in-class laboratory exercises during eight lecture periods, as well as extensive take-home research projects, which were intended to mimic the feeling of independent research by the students. In-class exercises were implemented using simulation software available to students through a cloud-based virtual desktop service. In addition to the in-class labs, students were given two group-based re- search projects in which they explored research problems requiring the use of these same concepts. The course chosen to test this method is an introductory junior level course in digital signal pro- cessing (DSP). This course was chosen because previous work has shown that hands-on experiences in DSP courses can increase students’ desire to learn (Adams, 2004). In addition, DSP laboratories are typically software based, making this course is a good choice for testing this approach because simulation software is available to students everywhere with an Internet connection through the cloud-based virtual desktop service. This work details the mixed learning approach and the assess- ment methods used to analyze the success of this approach. It is expected that this process will allow students to gain laboratory experience leading to the ability to complete more complex problems. The focus of this work is to describe a method for implementing laboratory experiences in a mid- year signal processing course, based on the Kolb cycle of experiential learning (Kolb 1984). This method may help signal processing educators implement experiential learning in their courses, and may be extended to other engineering courses. LITERATURE REVIEW In their review of the role of the laboratory in postsecondary education, Feisel and Rosa (2005) point out that the importance of laboratory-based learning has changed over time. They report that engineering programs in the early 19th century exposed students to both theory and practice. Then, in the mid-19th century, the focus turned largely to laboratory instruction, and this largely remained standard practice until the mid-20th century, when focus turned to educating engineers more in lec- ture halls than in laboratories or machine shops. Grayson (1993) explains that in the United States, this evolution of engineering education, can be understood in the context of the changing needs of the society from a newly founded pragmatic country, to one focused on westward expansion, to one focused more on the application of new science to solving problems. Today the importance of the laboratory as well as other hands-on design-centered activities has increased. Seeley (1999) points out that this is in recognition of the fact that many engineering students were completing engineering degrees without sufficient exposure to design and other hands-on experiences. As a result, engineering programs are grappling with how to fit laboratory components into an already crowded curriculum. SPRING 2017 3 ADVANCES IN ENGINEERING EDUCATION A Mixed Learning Approach to Integrating Digital Signal Processing Laboratory Exercises into a Non-Lab Junior Year DSP Course As mentioned previously, Lin and Tsai (2009) state that the learning that occurs in a laboratory setting can be considered to be at a higher level than in a lecture: students associate lectures with testing, calculating and practicing, while they feel that the laboratory gives them a chance to develop a deeper understanding of the relevant phenomena they need to learn to be effective engineers. One potential reason for a student’s deeper learning is their active engagement with the material in the lab. Another, as Lin and Tsai point out, is subtler: students see laboratory instructors as “supportive tutors” rather than a person who is “simply a lecturer.” With this being said, one might expect that students would want their education to focus on laboratories, but Lin and Tsai found that this was not the case – instead students generally prefer a mix of the two. In work published the same year as Lin and Tsai, Abdulwahed and Nagy (2009), argued that the traditional implementation of labs in engineering education (specifically in chemical engineering, which was their focus) was not effective. So, even if students prefer the mix, as argued by Lin and Tsai, the laboratory portion could be improved. Their comprehensive quantitative analysis was in the context of Kolb’s experiential learning cycle model, applied to the traditional laboratory. They define the traditional laboratory as a 3-hour meeting during which students accomplish an instructor- defined set of tasks. In the language of the Kolb theory, the traditional laboratory is mainly ineffec- tive because as implemented they do not require the student to sufficiently explore the prehension dimension of the learning cycle, which is to say: only reading the lab manual before doing the lab, does not set the stage for effective learning during and after lab. Abdulwahed and Nagy found that if the lab manual is supplemented by a “virtual pre-lab” the deficit in prehension is made up. They also suggest that the traditional 3-hour lab does not allow student sufficient time to reflect on the learning process (another dimension of the Kolb theory). There is a rich tradition of the incorporation of laboratory and hands-on components to DSP courses; see for example (McClellan 1998, Spanias 2005, Cameron 2014, and Mousa 2011). Among the varied topics in the engineering curriculum, DSP is uniquely suited to offer laboratory experi- ences that are outside of the traditional teaching laboratory. The main reason being that useful lab exercises can be designed so that they are computer-based, and not dependent on lab equipment that is not portable. Furthermore, as Adams and Mossayebi (2004) describe, students can actu- ally complete and extract valuable learning from some lab exercises in DSP without a complete knowledge of the material. They report that a number of DSP-related experiments were developed for a junior-level laboratory course, which students take before the DSP course at their institution. In another example of how laboratory experiences can be incorporated into coursework, Whitmal (2002) describes the use of a “studio” format of laboratory instruction in which class time is used for group-based laboratory experiences. Finally, although more specifically geared to mechanical instrumentation labs, Jouaneh and Palm (2009), describe how students can perform “take-home” 4 SPRING 2017 ADVANCES IN ENGINEERING EDUCATION A Mixed Learning Approach to Integrating Digital Signal Processing Laboratory Exercises into a Non-Lab Junior Year DSP Course labs, outside of class. In their paper, they describe the development of a carefully-designed hard- ware platform that the students use to complete the labs, something that would not be needed for a DSP-related lab, suggesting that it would be even easier to implement for DSP courses. MIXED LEARNING PROCESS As motivated by the literature review, there is a need for further research into the delivery and structure of laboratory material. In addition, as laboratory experiences are so vital for engineering students, it is important to accommodate labs into non-laboratory courses. There are a number of alternatives for the delivery of laboratory experiences that exist in literature and practice. A non- exhaustive list includes: 1. Scheduled laboratory periods 2. Holding out-of-class laboratory experiences, either during lecture or outside of scheduled lecture time (not a scheduled laboratory period) 3. Telecommunicated labs (Ogot, 2003) 4. “Living with the Lab” (Hall, 2008; Moller, 2015) 5. In class studio labs (Whitmal, 2002) 6. Large take-home projects (Jouaneh, 2009) Each of these solutions comes with a unique set of advantages and disadvantages. As outlined earlier, scheduled laboratory periods may not be possible due to credit limitations, instructor loading, or facility availability. The other alternatives accommodate for this fact by not requiring a scheduled lab period. Out-of-class laboratory experiences can be difficult to employ if students lack access to equipment or software needed for the laboratory experience. This is not the case for this particular implementa- tion of laboratory exercises for a signal processing course, as the students have access to MATLAB and Simulink through a cloud-based virtual desktop service. While telecommunicated labs are an option, even in a residential university, they are not consistent with the nature the engineering pro- gram in question, which places a high value on face-to-face meetings between faculty and students. Another method of delivering laboratory experiences is “Living with the Lab”, as implemented at Louisiana Tech University. In this method, students own their own tools, devices, and programmable controller, so they are able to implement labs at home, on their own time. While this is interesting, having students purchase the software needed for DSP is not needed in this case, as students have access to a cloud based virtual desktop service, which allows them to use MATLAB. However, some of the same principles can be applied to this course, allowing students to work on laboratory experiences at home. SPRING 2017 5 ADVANCES IN ENGINEERING EDUCATION A Mixed Learning Approach to Integrating Digital Signal Processing Laboratory Exercises into a Non-Lab Junior Year DSP Course The two remaining solutions are to incorporate laboratory exercises with lecture in a ‘studio’ format, or to give students ‘take-home’ laboratory experiments or projects. The studio format re- quires students to apply lecture concepts directly to in-class software based exercises (Whitmal, 2002). This method allows for students to learn engineering software applications of theoretical information with easy access to the instructor. This does take away from lecture time, and lacks time for student reflection. The take-home format involves assigning projects for students to complete outside of class, either alone or in a group (Jouaneh, 2009). This category of laboratory experience allows time for student reflection, but does not allow easy access to the instructor. As a result, some combination of the laboratory instruction alternatives was explored. The method used in this work was to combine the studio laboratory approach with the take-home project ap- proach. The reason for choosing these approaches was that they seemed the most suitable for signal processing, for the reasons listed with each alternative described. The solution described is called the mixed-learning approach, and has its roots in the cycle of experiential learning, as previously explained discussed in the literature review (Kolb, 1984). The foundation of the mixed learning approach is that students are able to develop a deep understanding of skills through an iterative process of learning, applying, and refining. The mixed learning process can be visualized by the flow chart in Figure 1. The cycle formed by this approach is similar to Kolb’s learning cycle. Kolb proposed that the optimal learning would be achieved through a cycle of Concrete Experience, Reflective Observation, Abstract Conceptualization, and Active Experimentation (Kolb, 1984; Abdulwahed, 2009). Instead of describing this cycle in Kolb’s vocabulary, it is presented here translated into terms more commonly used in engineering. The process starts with skills being introduced to students Figure 1. Mixed learning laboratory approach. 6 SPRING 2017 ADVANCES IN ENGINEERING EDUCATION A Mixed Learning Approach to Integrating Digital Signal Processing Laboratory Exercises into a Non-Lab Junior Year DSP Course through in-class studio labs. The skills that students develop independently in these studio labs are then applied to collaborative take-home projects, where they are further refined, and students are able to reflect on what they have learned. The refined skills can be applied back to studio lab assignments, resulting in students being able to learn more complex skills. Ideally, this process re- peats until students approach high proficiency in necessary skills, or alternatively, are able to gain moderate proficiency in more difficult skills. This cycle is informed by Kolb’s learning cycle in the following way: in-class studio labs provide concrete experience, and the outcome of the completion of these labs provides an opportunity for reflective observation. By applying learned skills to take home projects, students are asked to perform abstract conceptualization and active experimentation, which can lead to more impactful concrete experiences. In this way, the mixed learning approach is an application of Kolb’s cycle to laboratory experiences. Relating this approach to the course content, the first four studio labs are dedicated to teaching students skills necessary for completion of the first project. The students then apply these skills to the first project, which allows them to research and develop new methods by using these skills. These skills can be built upon to help students perform better in the final four studio labs, which are then applied to a more complex problem in Project 2. The specific learning outcomes for the signal processing laboratory experience are that 1. Students will be able to use MATLAB and Simulink to perform discrete signal manipulation, filtering, frequency analysis of discrete signals, and image processing. 2. Students will be able to communicate the results of their application of software skills to a technical audience. The achievement of these learning outcomes is assessed by blind review of student research papers using a rubric designed to measure these outcomes. The studio labs are designed to cover some of the most essential DSP concepts that show stu- dents hands-on applications of signal processing while reinforcing theory learned in the lecture component of the course. The number of labs is constrained to eight, as implementing labs in this way reduces the number of lectures that can be offered to cover theoretical material. This is less than a typical laboratory course, which typically involves 13-14 lab experiments. Because of the limited number of studio labs, take-home lab projects are a necessity. A detailed breakdown of the studio labs material is shown in Table 1. A number of topics were identified as essential based on reported work in other DSP courses (Adams, 2004; Ossman, 2008). The identified essential topics include: • Discrete signal manipulation • Filtering • Frequency analysis of discrete signals • Image processing SPRING 2017 7 ADVANCES IN ENGINEERING EDUCATION A Mixed Learning Approach to Integrating Digital Signal Processing Laboratory Exercises into a Non-Lab Junior Year DSP Course Table 1. Studio Lab Experiments. Studio Lab Description Lab 1 Discrete Signals and Systems Lab 2 Audio Processing Lab 3 Fast Fourier Transform Lab 4 Filter Design Lab 5 Point/Area Image Processing Lab 6 Blob Detection Lab 7 Edge Detection Lab 8 Equalization & Watermarking In-class labs are implemented during the scheduled lecture time, following the conclusion of the theoretical background necessary for each topic. This allows for continuity between the absorbed knowledge and applied knowledge, and ensures that students have the theoretical background fresh in their minds while developing hands-on skills. Students bring their laptops to class for the studio style labs. There are a number of benefits to students having laptops in class (Kolar, 2002) and the cloud based virtual desktop service provides students access to laboratory materials anywhere they have a high-speed Internet connection, and allows students with a variety of operating systems and devices to run engineering software. It is also notable that these studio labs could be delivered remotely via the Internet due to the accessibility of software. The studio laboratory experiments are augmented by two large research projects that require students to further refine skills learned in studio labs. As undergraduate research and collabora- tive assignments and projects are documented as high-impact educational practices, the project provides significant value to students (Kuh, 2008). The key aspects in designing the take-home projects was that the projects require an application of information from in-class studio labs, and that students be allowed appropriate time for reflection. As a result, the students were provided with research questions, but the execution and exploration of those questions were left open ended. The first project was centered on audio signal processing, as this requires manipulation of discrete signals, applications of filtering, and frequency analysis (McPheron, 2015a). The second project was an application of image processing, for which students had to manipulate of discrete signals, apply filters, and extend these skills to image processing. The first project required students to apply signal manipulation skills learned in the first four studio labs to the development of a novel reverberation algorithm for audio signals. A simple reverberation algorithm is shown in Figure 2 (McPheron, 2015b). Students chose to explore a number of topics related to reverberation including the use of chorus modulation, vibrato, and wavelet compression techniques. The technical steps associated with this project are: 8 SPRING 2017 ADVANCES IN ENGINEERING EDUCATION A Mixed Learning Approach to Integrating Digital Signal Processing Laboratory Exercises into a Non-Lab Junior Year DSP Course Figure 2. Simple Reverberation. 1. Research algorithms 2. Develop a novel algorithm or area to study 3. Implement reverb algorithm in MATLAB and Simulink 4. Test algorithms 5. Iterate until desired result The second project required students to apply image processing skills developed in the last four studio labs to tracking multiple objects through video (McPheron, 2014). A still frame from the video used is shown in Figure 3 (McPheron, 2015b). Students were asked to further develop a Figure 3. Still frame for multiple-identical blob tracking. SPRING 2017 9 ADVANCES IN ENGINEERING EDUCATION A Mixed Learning Approach to Integrating Digital Signal Processing Laboratory Exercises into a Non-Lab Junior Year DSP Course tracking algorithm designed to follow multiple blobs through the video. This project is significantly more complex than the first project, requiring 8 technical steps, opposed to the 5 steps needed for Project 1. These steps are: 1. Research algorithms 2. Develop a novel algorithm or area to study 3. Get frames from video 4. Extract blobs 5. Suppress extra blobs 6. Implement tracking algorithm in MATLAB and Simulink 7. Test algorithms 8. Iterate until desired result For these projects, students were split into small multidisciplinary teams of 2 or 3 students for work on the projects presented in this paper. The overall student population was composed of 11 students from disciplines including Electrical Engineering, Computer Engineering, and Computer Science. Each project begins with a proposal phase, in which students propose a topic for study in the form of an abstract. The instructor is responsible for guiding the students in their research by ap- proving or revising the abstract and providing constructive feedback meetings with each group. The main deliverable for each project is a technical report, written in IEEE standard format, which details the theory and methods used in the project. In addition to the required technical report, students are responsible for submitting any code written for the project, in order to verify the skills that the students have applied and confirm that students have prepared their own assignments. ASSESSMENT To assess the effectiveness of this method, a direct assessment tool is used. Although student surveys can be used as an indirect measure of the method, the student’s perceived understanding is often different than actual performance (Bernadin, 2007). As a direct method of assessment, student-project technical papers are submitted to several faculty members for blind review. These faculty members rate student ability using a prescriptive rubric, shown in Table 2. The rubric was designed with two purposes in mind. The first, and most important, is that the rubric assesses the achievement of the learning outcomes for the laboratory experience. This means that student completion of technical tasks and ability to communicate their results should be measured. The second purpose is to expose students to a review process similar to a conference. Many conferences possess unique rubrics for reviewer decision, which aided in 10 SPRING 2017