Pre-Conference Talk by HU Changshuo | From Cheap to Chic : Enhancing Music Playback Quality of Budget Earphones via Hardware-Aware Learning

Low-end earphones are widely spread due to their affordability, but their limited speaker hardware often leads to poor music playback quality. This raises a key question: Can we compensate for hardware limitations to enhance the listening experience without modifying the device? Existing EQ-based approaches attempt this, but they rely on frequency response curves (FRCs) measured under ideal conditions, which fail to capture real-world distortions such as harmonic and intermodulation effects. To address this, we propose Cheap2Chic, a two-stage, speaker-aware framework that improves music playback quality on low-end earphones using deep learning. In stage one, Cheap2Chic trains a speaker characterization model to capture the playback properties of the target earphone. In stage two, it introduces a digital audio enhancement model, prepended before the frozen speaker characterization model, that learns to adapt the input audio to compensate for hardware-induced distortions. To reduce the data collection burden required for accurate speaker characterization, Cheap2Chic also incorporates a data synthesis module. We built a prototype and collected a dataset to evaluate Cheap2Chic under diverse real-world conditions. Results show that Cheap2Chic significantly improves perceived music quality both objectively and subjectively, while remaining efficient on mobile devices.

This is a Pre-Conference talk for ACM/IEEE International Conference on Embedded Artificial Intelligence and Sensing Systems (SenSys 2026).

Changshuo Hu is currently a Ph.D. candidate in Computer Science at Singapore Management University, under the supervision of Professor Archan Misra, working closely with Professor Dong Ma at the University of Cambridge. His research interests encompass mobile computing and pervasive sensing, with a particular focus on human-centric sensing and interaction through earable devices, spanning authentication, physiological monitoring, and audio-driven system optimization.