Speech2Grasp: Data-Efficient Transfer of Text-Conditioned Grasp Detection to Speech in Humanoid Robots

2026-07-29

AI Summary

This work addresses the challenge of enabling humanoid robots to directly execute grasping tasks from spoken commands, circumventing the limitations of existing approaches that either rely on text-based vision-language models or cascade automatic speech recognition (ASR) with downstream modules—introducing both latency and error propagation. To this end, the authors propose Speech2Grasp, a framework that employs a lightweight MLP projector to align speech features with the semantic space of a pretrained text-conditioned grasping model built upon ALBEF. This design facilitates efficient cross-modal transfer, allowing adaptation from text to speech with only minimal training data. Real-robot experiments demonstrate that Speech2Grasp significantly outperforms conventional ASR-cascaded pipelines, achieving higher grasping accuracy and lower inference latency.

Abstract

Humanoid robots increasingly require multi-modal understanding for natural interaction with humans. Despite the prominence of vision-language models, they generally assume textual rather than the more natural speech inputs. In this paper, we investigate whether a well-established text-conditioned model can be transferred to speech in a data-efficient manner. Using ALBEF as a case study, we conduct diagnostic analyses showing that a lightweight MLP-based projector effectively adapts it to speech, while preserving semantic discrimination and robustness. Motivated by these findings, we introduce Speech2Grasp, a framework for data-efficient transfer of text-conditioned grasp detection to speech. Real-world humanoid robot experiments show that Speech2Grasp outperforms cascaded ASR-based pipeline, while reducing inference latency. Our findings suggest a practical paradigm for extending established text-conditioned systems to speech.

Problem

• speech-conditioned grasp detection
• data-efficient transfer
• humanoid robots
• text-to-speech adaptation
• multi-modal interaction

Innovation

• speech-conditioned grasping
• data-efficient transfer
• lightweight MLP projector
• vision-language model adaptation
• humanoid robot interaction