The present invention discloses an audio speaker recognition method, system, storage medium, and electronic device, belonging to the fields of audio processing and artificial intelligence. The method comprises: preprocessing the input audio, extracting channel information, and determining whether the audio is mono or stereo; for stereo audio, determining whether it is pseudo‑stereo (dual‑mono) or true stereo by comparing at least two acoustic feature parameters between the left and right channels; selecting a processing strategy based on the audio type; sequentially performing, on the selected channel(s), denoising preprocessing, voice activity detection (VAD), speaker diarization (segmentation), speech recognition, and punctuation restoration; using a large language model to assign speaker‑role labels to the recognized content; and merging content segments belonging to the same speaker role to generate structured output. Through the cooperation of multi‑dimensional acoustic feature analysis, deep learning models, and large language models, the invention improves recognition accuracy and the degree of automation, and can adapt to various complex audio scenarios.
还没有评价,成为第一个了解它的人。
Copyright © 2026 RADHOST PTE. LTD. (UEN: 202450723N). All Rights Reserved.