Spatial Audio Processing with Large Language Model on Wearable Devices
Ayushi Mishra, Yang Bai, Priyadarshan Narayanasamy, Nakul Garg, Nirupam Roy
TL;DR
This work addresses the challenge of enabling spatial speech understanding within large language models on wearable devices. It introduces SING, a framework that combines a monaural Owlet-inspired microstructure with a spatial encoder and Whisper-based linguistic embeddings, aligned to an LLM via a lightweight projection and LoRA fine-tuning. Key contributions include the OmniTalk synthetic dataset, a DoA-aware spatial speech encoder, and a multi-DoA soundscaping module, achieving $MAE=25.72^\ 0$ and $WER=5.3\%$ on spatial ASR, with robust performance across up to five simultaneous speakers. The results demonstrate the feasibility of on-device, privacy-preserving spatially aware LLMs for wearables, enabling applications in AR, accessibility, and immersive experiences, while outlining avenues for future 3D localization, real-world data, and multimodal integration.
Abstract
Integrating spatial context into large language models (LLMs) has the potential to revolutionize human-computer interaction, particularly in wearable devices. In this work, we present a novel system architecture that incorporates spatial speech understanding into LLMs, enabling contextually aware and adaptive applications for wearable technologies. Our approach leverages microstructure-based spatial sensing to extract precise Direction of Arrival (DoA) information using a monaural microphone. To address the lack of existing dataset for microstructure-assisted speech recordings, we synthetically create a dataset called OmniTalk by using the LibriSpeech dataset. This spatial information is fused with linguistic embeddings from OpenAI's Whisper model, allowing each modality to learn complementary contextual representations. The fused embeddings are aligned with the input space of LLaMA-3.2 3B model and fine-tuned with lightweight adaptation technique LoRA to optimize for on-device processing. SING supports spatially-aware automatic speech recognition (ASR), achieving a mean error of $25.72^\circ$-a substantial improvement compared to the 88.52$^\circ$ median error in existing work-with a word error rate (WER) of 5.3. SING also supports soundscaping, for example, inference how many people were talking and their directions, with up to 5 people and a median DoA error of 16$^\circ$. Our system demonstrates superior performance in spatial speech understanding while addressing the challenges of power efficiency, privacy, and hardware constraints, paving the way for advanced applications in augmented reality, accessibility, and immersive experiences.