RevComm Research Paper on 'Automatic Estimation of Speaker Diarization Error Rate' Accepted at International Conference ICASSP 2026

RevComm Research Paper Accepted at ICASSP 2026

A research paper by RevComm Research (RCR), the R&D division of RevComm, on technology to predict whether an AI can accurately interpret audio, has been accepted at ICASSP 2026 (Barcelona, Spain), one of the world's largest international conferences on voice and acoustic signal processing, held from May 4 to 8, 2026.

### About ICASSP

ICASSP (International Conference on Acoustics, Speech, and Signal Processing) is hosted by the IEEE Signal Processing Society, which holds the longest history in signal processing within the IEEE.

### Research Content

The paper, "Automatic Estimation Of Speaker Diarization Error Rate Based On Features Of Audio Quality And Speaker Discriminability," authored by Senior Research Engineers Kenkichi Ishizuka and Masaki Ohno, and Research Director Taiichi Hashimoto of RCR, has been accepted.

While voice AI applications are expanding, their accuracy often depends on the state of the recorded audio. AI may fail to analyze audio correctly in cases of high background noise, overlapping speech, or indistinguishable voices. This study proposes a method to automatically determine if the audio is 'easy for AI to parse' by analyzing features such as audio quality and speaker discriminability, and predicting the error rate of speaker diarization in advance. Experimental results confirmed a high correlation between the predicted and actual error rates, validating the method.

This research allows for better distinction between AI model issues and environmental factors, expected to improve voice AI service quality and serve as a metric for improving recording environments.

- Title: Automatic Estimation Of Speaker Diarization Error Rate Based On Features Of Audio Quality And Speaker Discriminability - URL: https://ieeexplore.ieee.org/document/11463892

### About RevComm Research (RCR)

RCR is an organization dedicated to solving communication challenges using voice technology and AI, based on the corporate philosophy: 'Reinventing communication and creating a society where people care for one another.'

- RCR Website: https://www.revcomm.co.jp/rcr/