Abstract
The extraction of speech emotion features—particularly at the utterance level—constitutes a critical yet challenging aspect of Speech Emotion Recognition (SER). Emotion represents high-level paralinguistic information, and not all segments within a speech signal carry emotionally salient cues, especially in longer utterances. Conventional methods that process entire utterances may thus waste computational resources and introduce irrelevant acoustic interference. To address this, we propose AdaPre-Selection, an adaptive pre-selection mechanism designed to identify and extract emotion-rich segments from speech signals. Acting as a flexible front-end compression module, AdaPre-Selection consists of two complementary components: an Active Emotion Positioning (AEP) module and a Passive Emotion Constraint (PEC) module. Within AEP, Temporal Length Selection (TLS) and Start Time Point Selection (STPS) operate jointly to adaptively locate the optimal emotional segment in each utterance. Evaluated on two benchmark datasets (IEMOCAP and MSP-IMPROV) using four state-of-the-art SER models, AdaPre-Selection consistently outperforms common preprocessing baselines and delivers the most substantial improvement in recognition performance.
Funding Information
  • Science and Technology Research Project of Henan Province (252102320055)
  • Science and Technology Research Project of Henan Province (252102310178)
  • Science and Technology Research Project of Henan Province (252102210100)
  • Science and Technology Research Project of Henan Province (262102220025)
  • Education Department of Henan Province (26A520012)
  • Education Department of Henan Province (26A520013)
  • Henan Province Natural Science Foundation Young Scientists Fund (252300423320)
  • Henan Province Natural Science Foundation Young Scientists Fund (252300423775)
  • Science and Technology Research Project of Jiaozuo, Henan Province (2025210097)

This publication has 31 references indexed in Scilit: