Research Experts

Top Conference

2026

Ten papers accepted to the international conference ICASSP2026

Ten papers from the Mitsubishi Electric Group were accepted at the 2026 IEEE International Conference on Acoustics, Speech, and Signal Processing, which is a leading international conference in this field.
Additionally, one paper was accepted for a workshop at the same conference, and two papers published in a journal will also be presented (Listed below).
ICASSP2026 will be held in Barcelona, Spain from May 4 to May 8, 2026.

2026 IEEE International Conference on Acoustics, Speech, and Signal Processing (External site)(open new window)

Title
Low-Computation Detection Method for Unknown LFM Signals below Noise Floor
Authors
Hirofumi Fukushima (Melco), Yuki Takabayashi (Melco)
URL
LOW-COMPUTATION DETECTION METHOD FOR UNKNOWN LFM SIGNALS BELOW NOISE FLOOR (External site)(open new window)
Title
One-Stage Semi-Supervised Semantic Segmentation for Anomaly Detection via Consistency Regularization and Student-Teacher Models
Authors
Yuta Nakanishi (Melco)
URL
ONE-STAGE SEMI-SUPERVISED SEMANTIC SEGMENTATION FOR ANOMALY DETECTION VIA CONSISTENCY REGULARIZATION AND STUDENT-TEACHER MODELS (External site)(open new window)
Title
Dual-Regularized Iterative Adaptive Approach for DOA Spectrum Reconstruction in Limited Angle Sector
Authors
Ryuhei Takahashi (Melco), Hassan Mansour (MERL), Petros Boufounos (MERL)
URL
DUAL-REGULARIZED ITERATIVE ADAPTIVE APPROACH FOR DOA SPECTRUM RECONSTRUCTION IN LIMITED ANGLE SECTOR (External site)(open new window)
Title
SUNAC: Source-aware Unified Neural Audio Codec
Authors
Ryo Aihara (Melco/MERL), Yoshiki Masuyama (MERL), Francesco Paissan (Univeristy of Trento), François G. Germain (MERL), Gordon Wichern (MERL), Jonathan Le Roux (MERL)
URL
SUNAC: Source-aware Unified Neural Audio Codec (External site)(open new window)
Title
FlexIO: Flexible Single- and Multi-Channel Speech Separation and Enhancement
Authors
Yoshiki Masuyama (MERL), Kohei Saijo (Waseda University), Francesco Paissan (Univeristy of Trento), Jiangyu Han (Brno University of Technology), Marc Delcroix (NTT, Inc.), Ryo Aihara (Melco/MERL), François G. Germain (MERL), Gordon Wichern (MERL), Jonathan Le Roux (MERL)
URL
FLEXIO: FLEXIBLE SINGLE- AND MULTI-CHANNEL SPEECH SEPARATION AND ENHANCEMENT (External site)(open new window)
Title
Velocity Potential Neural Field for Efficient Ambisonics Impulse Response Modeling
Authors
Yoshiki Masuyama (MERL), François G. Germain (MERL), Gordon Wichern (MERL), Chiori Hori (MERL), Jonathan Le Roux (MERL)
URL
VELOCITY POTENTIAL NEURAL FIELD FOR EFFICIENT AMBISONICS IMPULSE RESPONSE MODELING (External site)(open new window)
Title
Spatially Aware Self-Supervised Models for Multi-Channel Neural Speaker Diarization
Authors
Jiangyu Han (Brno University of Technology), Ruoyu Wang (University of Science and Technology of China), Yoshiki Masuyama (MERL), Marc Delcroix (NTT, Inc.), Johan Rohdin (Brno University of Technology), Jun Du (University of Science and Technology of China), Lukas Burget (Brno University of Technology)
URL
SPATIALLY AWARE SELF-SUPERVISED MODELS FOR MULTI-CHANNEL NEURAL SPEAKER DIARIZATION (External site)(open new window)
Title
Heatmap-to-SMPL Multi-View Radar Transformer for Multi-Person 3D Pose Estimation
Authors
Sorachi Kato (University of Osaka), Pu (Perry) Wang (MERL), Takuya Fujihashi (University of Osaka), Andrew Markham (University of Oxford)
URL
Heatmap-to-SMPL Multi-View Radar Transformer for Multi-Person 3D Pose Estimation (External site)(open new window)
Title
Single View Camera-Based Dynamic Airflow Sensing
Authors
Kevin Tandi (University of California San Diego), Wael H. Ali (MERL), Joshua Rapp (MERL), Hassan Mansour (MERL)
URL
SINGLE VIEW CAMERA-BASED DYNAMIC AIRFLOW SENSING (External site)(open new window)
Title
ProxiCBO: A Consensus-based Method for Composite Optimization
Authors
Haoyu Zhang (University of California San Diego), Yanting Ma (MERL), Ruangrawee Kitichotkul (Boston University), Joshua Rapp (MERL), Petros T. Boufounos (MERL)
URL
PROXICBO: A CONSENSUS-BASED METHOD FOR COMPOSITE OPTIMIZATION (External site)(open new window)

List of papers from the Mitsubishi Electric Group accepted for the workshop

One paper was accepted for the workshop (Low-Resource Audio Codec (LRAC) Workshop) at the same conference.

The 2026 LRAC Workshop (External site)(open new window)

Title
Exploring Disentangled Neural Speech Codecs from Self-Supervised Representations
Authors
Ryo Aihara (Melco/MERL), Yoshiki Masuyama (MERL), Gordon Wichern (MERL), François G. Germain (MERL), Jonathan Le Roux (MERL)
URL
Exploring Disentangled Neural Speech Codecs from Self-Supervised Representations (External site)(open new window)

List of papers from Mitsubishi Electric Group published in the journal and will be presented

Two papers published in the IEEE Open Journal of Signal Processing will also be presented at the same conference.

Journal
Embracing Cacophony: Explaining and Improving Random Mixing in Music Source Separation
IEEE Open Journal of Signal Processing, pp. 1179-1192, vol. 6, 2026
Authors
Chang-Bin Jeon (Samsung Electronics), Gordon Wichern (MERL), François G. Germain (MERL), Jonathan Le Roux (MERL)
URL
Embracing Cacophony: Explaining and Improving Random Mixing in Music Source Separation (External site)(open new window)
Journal
SuDaField: Subject- and Dataset-Aware Neural Field for HRTF Modeling
IEEE Open Journal of Signal Processing, pp. 1169-1178, vol. 6, 2026
Authors
Yoshiki Masuyama (MERL), Gordon Wichern (MERL), François G. Germain (MERL), Christopher Ick (New York University), Jonathan Le Roux (MERL)
URL
SUDAFIELD: SUBJECT- AND DATASET-AWARE NEURAL FIELD FOR HRTF MODELING (External site)(open new window)

Three papers accepted at the international conference AAAI-26

Three papers from the Mitsubishi Electric Group were accepted at the 40th Annual AAAI Conference on Artificial Intelligence (AAAI-26), a leading international conference on artificial intelligence.
Additionally, five papers were accepted for the workshops at the same conference (List publication).
AAAI-26 will be held in Singapore from January 20 to January 27, 2026.

Summary of accepted three papersPDF

Title
Indoor Multi-View Radar Object Detection via 3D Bounding Box Diffusion
Authors
Ryoma Yataka (Melco), Pu (Perry) Wang (MERL), Petros Boufounos (MERL), Ryuhei Takahashi (Melco)
Title
LatentLLM: Activation-Aware Transform to Multi-Head Latent Attention
Authors
Toshiaki Koike-Akino (MERL), Xiangyu Chen (MERL), Jing Liu (MERL), Ye Wang (MERL), Pu (Perry) Wang (MERL), Matthew Brand (MERL)
Title
Chain-of-Thought Driven Adversarial Scenario Extrapolation for Robust Language Models
Authors
Md Rafi Ur Rashid (Pennsylvania State University), Vishnu Asutosh Dasu (Pennsylvania State University), Ye Wang (MERL), Gang Tan (Pennsylvania State University), Shagufta Mehnaz (Pennsylvania State University)

Papers from the Mitsubishi Electric Group Accepted at Workshops in 2026

At the workshops of the same conference, we will present five papers.

Title
Amplification Effects in Test-Time Reinforcement Learning: Safety and Reasoning Vulnerabilities
Authors
Vanshaj Khattar (Virginia Tech University), Moumita Choudhury (The University of Massachusetts), Md Rafi Ur Rashid (Pennsylvania State University), Jing Liu (MERL), Toshiaki Koike-Akino (MERL), Ming Jin (Virginia Tech University), Ye Wang (MERL)
Title
PatchDecomp: Interpretable Patch-Based Time Series Forecasting
Authors
Hiroki Tomioka (Melco), Genta Yoshimura (Melco)
Title
Perception-Reasoning Integration for Temporal Action Segmentation in Human-Centered Manufacturing
Authors
Jia Qu (Melco), Ren Sasage (Melco), Shotaro Miwa (Melco)
Title
A Task-Level Explanation Framework for Meta-Learning Algorithms
Authors
Yoshihiro Mitsuka (Melco), Shadan Golestan (Alberta Machine Intelligence Institute), Zahin Sufiyan (University of Alberta), Shotaro Miwa (Melco), Osmar Zaiane (University of Alberta, Alberta Machine Intelligence Institute)
Title
GeoSteer: Faithful Chain-of-Thought Steering via Latent Manifold Gradients
Authors
Kentaro Kazama (Melco), Daiki Shirafuji (Melco), Tatsuhiko Saito (Melco)

Return to "Research Experts" Top Page