IN FOCUS: Dr. Leslie Collins, IEEE Fellow
By Joanne Van Voorhis
Dr. Leslie Collins is a Professor of Electrical and Computer Engineering at Duke University, where she also holds an appointment in the Department of Biomedical Engineering and the Department of Head and Neck Surgery & Communication Sciences, and directs the Applied Machine Learning Lab. She is interested in developing physics-based machine learning and AI algorithms, including developing remediation strategies for the hearing impaired and sensor-based algorithms for the detection of hazardous buried objects such as land mines. Dr. Collins was elected an IEEE Fellow in 2024 for contributions to signal processing algorithms for auditory applications and to buried threat detection.
Improving Mine Detection Accuracy…and Beyond
Related to mine detection, her work has helped reduce false positives for those searching for underground threats by increasing the accuracy and specificity of the incoming signal. Metal detectors measure the energy given off by metal buried in the ground, but usually in the process of creating an alert, the individual signals of the metals merge into one – making a mine indistinguishable from a harmless piece of metal such as a nail or other metal fragment. Collins’ work has greatly improved the efficiency of mine identification globally by boosting detection accuracy and reducing false alarms.
Dr. Collins has pursued other remote sensing applications, including detection and classification of tanks, cars, people, and clutter items in IR imagery, detection of energy infrastructure and buildings in satellite imagery. She has also been involved in a variety of biomedical research applications including brain computer interfaces, seizure detection, traumatic brain injury classification and heart sound classification.
Impacting the Lives of the Hearing Impaired
Dr. Collins is also recognized for her work on cochlear implants, which are electronic devices that provide a sense of hearing and the ability to recognize speech for people with severe hearing loss. They work by bypassing the damaged inner ear (cochlea) and directly stimulating the auditory nerve. Cochlear implant recipients often struggle to understand speech in reverberant environments. Collins has focused on improved speech enhancement algorithms that could restore speech perception for CI listeners by removing reverberant artifacts from the CI stimulation pattern. Human listening studies are costly and time-consuming during algorithm development, so Collins has focused on estimating speech recognition using objective intelligibility measures in reverberant-only conditions characterized by a gradual reduction of reverberant artifacts to achieve boosted speech recognition. Her most recent work in this area has been in leveraging advanced AI approaches to reverberation mitigation.
Dr. Collins became interested in using advanced signal processing techniques for improving speech recognition during her PhD studies at the University of Michigan, Ann Arbor. She can still clearly recall how excited she was when one of the human experimental subjects exclaimed “Wow! That sounds so much clearer” when she turned their device on using the algorithm that she had developed. As with many researchers, knowing that your research has a real impact, be it generating clearer speech or finding landmines more accurately, is a strong motivator.
Dedicated Duke Professor and Colleague
Collins earned her bachelor’s degree in Electrical Engineering from the University of Kentucky, and her master’s degree from the University of Michigan. She worked as an engineer for Westinghouse Electric Corporation for five years and then returned to the University of Michigan to earn her PhD. She has been a faculty member at Duke University since 1995, initially as an assistant professor. She was tenured as an associate professor in 2002 and promoted to full professor in 2007.
Dr. Collins enjoys almost all aspects of university life – teaching undergraduates the fundamentals of systems and systems, teaching graduate students advanced machine learning techniques, and mentoring both undergraduate and graduate students in her lab. “Duke has provided me with an exceptionally supportive and inclusive environment with significant academic freedom,” she explains. “I have developed some invaluable colleagues not only within Engineering, but also in the Medical School, the School of the Environment, and the Sanford Institute.”
The Success of Interdisciplinary Collaboration
Dr. Collins frequently collaborates with colleagues within Duke and beyond. For example, collaborating with the Duke School of Medicine, Dr. Collins has developed machine learning algorithms that analyze acoustic signals from digital stethoscopes to detect complications in patients with Left Ventricular Assist Devices (LVADs). This innovative approach allows for early detection of issues such as blood clots, potentially preventing serious health events. Dr. Collins’ interdisciplinary approach, combining engineering principles with clinical insights, has significantly contributed to the advancement of patient care for individuals with LVADs. Her work exemplifies the potential of integrating technology and medicine to address complex healthcare challenges.
Early Student Research Experiences
Collins encourages students to get involved in research experiences early to help them decide which career path they prefer – academia or industry. A good example is Duke Data+, an immersive, interdisciplinary summer program designed to foster student growth through hands-on learning, research, and collaboration. Dr. Collins has made significant contributions to the Duke Data+ Program, a ten-week summer research experience for undergraduates and master’s students interested in data-driven interdisciplinary projects. In 2015, in collaboration with the Energy Data Analytics Lab, Dr. Collins mentored a team analyzing satellite imagery to identify solar panels using image recognition. The project successfully developed a proof-of-principle algorithm with over 90% accuracy in detecting solar panels, contributing to advancements in renewable energy analytics. In 2022, she led a team exploring novel machine learning techniques for Brain-Computer Interface (BCI) applications. This project aimed to develop innovative methods to enhance communication between the brain and external devices, reflecting her expertise in signal processing and neural interfaces. “These research experiences can make a significant impact on student growth and help them thoughtfully consider career options too,” Collins explains.
Career Moments that Matter
Dr. Collins has experienced many high points in her career…from winning algorithm competitions to hooding PhD students. “It’s a journey,” she explains, “and that PhD hood is a clear demonstration of a dedicated student transitioning to the next step in a career path.” “I especially love witnessing the lightbulb come on – whether it’s an undergraduate suddenly grasping convolution or Fourier analysis, or a PhD student achieving a breakthrough with an algorithm, each spark of understanding is equally exciting,” she adds.
Conferences as Catalysts for Growth
Dr. Collins is a big supporter of attending conferences and continuing to learn from colleagues around the world. “Being a member of IEEE, GRSS, and other professional societies allows me to stay current with the latest research through journals, while conferences provide not only valuable learning opportunities but also the chance to connect with people – many of whom have become long-term colleagues,” she explains. In fact, if she could have a chat with her 22-year-old self, she would advise going to more conferences. “Conference attendance became much more difficult, and I felt like I fell behind, when I had children. Not that I would trade them for anything, but it certainly complicated my professional life.”
Find out more about Dr. Collins and her research at the following links:
