Apple is set to showcase significant advancements in artificial intelligence and machine learning at the annual International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2026. The event, a crucial gathering point for scientific and industrial research communities focused on signal processing, will take place in person in Barcelona, Spain, from May 4 to 8. Apple proudly sponsors this conference and will maintain an active presence with booth #P2 at the Centre de Convencions Internacional de Barcelona (CCIB). Attendees are invited to visit the booth during exhibition hours: Monday, May 4, from 19:00 to 21:30 CEST, and from Tuesday, May 5, through Friday, May 8, between 09:00 and 17:00 CEST.
Among the key research contributions, Apple will present a poster titled "Leveraging Audio — Visual Data to Reduce the Multilingual Gap in Self — Supervised Speech Models." Co-authored by researchers including María Andrea Cruz Blandón from Tampere University, Zakaria Aldeneh, Jie Chi, and Maureen de Seyssel, this work in Speech and Natural Language Processing seeks to enhance the inclusivity and performance of speech technologies across diverse languages. The poster session is scheduled for Wednesday, May 6, from 16:30 to 18:30 CEST, in Poster Area 27 as part of the SLP — P25: Audio — Visual Speech Recognition session. This research represents a significant step towards developing more universally accessible voice interfaces.
Another notable presentation is the poster "StereoFoley: Object — Aware Stereo Audio Generation from Video." This collaborative research, involving Tornike Karchkhadze from UC San Diego, Kuan — Lin Chen, Mojtaba Heydari, Robert Henzel, Alessandro Toso, Mehrez Souden, and Joshua Atkins, falls under the Tools, Platforms, Frameworks area. It focuses on generating realistic stereo audio that is contextually aware of objects within a video, an innovation with profound implications for creating more immersive virtual and augmented reality experiences. The poster will be displayed on Friday, May 8, from 14:00 to 16:00 CEST, in Poster Area 25 during the AASP — P30: Audio for Video and Multimedia session.
Apple's third significant contribution is an oral presentation titled "Principled Coarse — Grained Acceptance for Speculative Decoding in Speech." Authored by Moran Yanuka, Paul Dixon, Eyal Finkelshtein, Daniel Rotman, and Raja Giryes from Tel Aviv University, this paper addresses crucial aspects of efficient speech processing within the Methods and Algorithms and Speech and Natural Language Processing research areas. This work explores techniques to optimize speculative decoding, which could lead to substantial improvements in the speed and efficiency of text-to-speech (TTS) systems and other streaming speech applications, making interactive AI more responsive.
Beyond presenting research, Apple's commitment to advancing the field is demonstrated through its substantial contributions to the conference's organizational structure. Daniele Giacobello, Kumari Nishu, Nimshi Venkat Meripo, Sakshi Rathi, Sibel Oyman, Xuankai Chang, and Zijin Gu are serving as Area Chairs, guiding the peer-review process and shaping the scientific program. Bo — Hao Su is contributing as a Session Chair and, alongside Honor Chen, as a reviewer. Furthermore, Daniele Giacobello is a Workshop Co — Organizer for “Data Science for Telepresence and Extended Reality Applications 2026,” showcasing Apple's dedication to fostering scientific discourse and collaboration within the signal processing community.
These collective efforts at ICASSP 2026 underscore Apple's continuous dedication to pioneering advancements in machine learning and signal processing. The presented research, spanning from enhancing multilingual speech models and generating object — aware stereo audio to optimizing speech decoding, showcases a multifaceted approach. Such breakthroughs are instrumental in improving user experiences across Apple's product ecosystem and contribute significantly to the broader landscape of artificial intelligence.
Sources
Replies (0)
No replies in this topic yet.
