TellWhisper: Tell Whisper Who Speaks When

TL;DR

TellWhisper introduces a unified speech encoder that jointly models speaker identity and timing, improving multi-speaker speech recognition especially during rapid turn-taking and overlaps.

Contribution

It proposes TS-RoPE for explicit time-speaker encoding and Hyper-SD for hyperbolic speaker activity estimation, addressing limitations of previous decoupled methods.

Findings

Significantly improves recognition accuracy in overlapping speech scenarios.

Effectively captures speaker turn transitions and state dynamics.

Enhances inter-class separation in speaker activity estimation.

Abstract

Multi-speaker automatic speech recognition (MASR) aims to predict ''who spoke when and what'' from multi-speaker speech, a key technology for multi-party dialogue understanding. However, most existing approaches decouple temporal modeling and speaker modeling when addressing ''when'' and ''who'': some inject speaker cues before encoding (e.g., speaker masking), which can cause irreversible information loss; others fuse identity by mixing speaker posteriors after encoding, which may entangle acoustic content with speaker identity. This separation is brittle under rapid turn-taking and overlapping speech, often leading to degraded performance. To address these limitations, we propose TellWhisper, a unified framework that jointly models speaker identity and temporal within the speech encoder. Specifically, we design TS-RoPE, a time-speaker rotary positional encoding: time coordinates are derived from frame indices, while speaker coordinates are derived from speaker activity and pause cues. By applying region-specific rotation angles, the model explicitly captures per-speaker continuity, speaker-turn transitions, and state dynamics, enabling the attention mechanism to simultaneously attend to ''when'' and ''who''. Moreover, to estimate frame-level speaker activity, we develop Hyper-SD, which casts speaker classification in hyperbolic space to enhance inter-class separation and refine speaker-activity estimates. Extensive experiments demonstrate the effectiveness of the proposed approach.