The University of Southern Maine’s Computer Science program continues to gain recognition for its emphasis on undergraduate research and innovation, as faculty and students prepare to present their latest work at an international conference.
Yuqi Song and Xin Zhang, alongside three undergraduate computer science students, have completed a collaborative research project focused on improving the efficiency of automatic speech recognition (ASR) systems. Notably, the project is student-led, with an undergraduate serving as the first author and playing a central role in the development of the research.
The team’s work introduces a novel ASR model, called IBNet, which enhances computational efficiency by incorporating inverted bottleneck modules into a convolutional neural network architecture. This design allows the model to maintain strong performance while reducing computational demands, which is an important consideration in real-world speech processing applications.
To evaluate the model, the researchers conducted experiments using the widely recognized LibriSpeech dataset. Results demonstrated that IBNet achieves performance competitive with existing baseline models, highlighting its potential as an efficient alternative in ASR tasks.
The paper has been accepted for presentation at the 24th IEEE/ACIS International Conference on Software Engineering Research, Management and Applications (SERA 2026), an international venue that brings together researchers, practitioners, and students to share new ideas, research results, and practical solutions across all areas of software engineering. As part of this recognition, the student authors have been invited to present their findings at the conference, offering them valuable experience in sharing research on an international stage.
This achievement underscores the University of Southern Maine’s commitment to fostering hands-on learning, supporting undergraduate scholarship, and advancing applied research in artificial intelligence and computer science.