Unverified50% confidenceFactExact time
Mel频谱图基于心理声学研究，通过对数变换模拟人耳对低频比高频更敏感的非线性感知特性
1
Sources
50%
Confidence
Long-term
Relevance
7/20/2026
First Seen
Sources
Related Claims
UnverifiedThe Mel spectrogram simulates the nonlinear perceptual characteristics of the human ear across different frequencies, converting raw audio waveforms into a two-dimensional time-frequency representation that is convenient for convolutional neural networks to perform pattern recognition.77% similarUnverifiedMel频谱图是模拟人耳听觉感知非线性特性的时频表示，是Whisper等端到端模型的标准输入格式71% similarUnverified科林·切里通过二分听觉（Dichotic Listening）实验发现人类大脑能依靠音调、音色、空间方位和语义线索选择性关注特定声源68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/573856API
curl https://kongchang.com/api/v1/knowledge/claims/573856MCP
get_claim(id=573856)