Unverified50% confidenceFactExact time
HMM假设语音波形由音素状态序列依概率生成，解码时通过维特比算法寻找后验概率最大的状态路径，并通常与高斯混合模型结合形成GMM-HMM架构
1
Sources
50%
Confidence
Long-term
Relevance
7/21/2026
First Seen
Sources
Related Claims
UnverifiedASR技术最初基于隐马尔可夫模型（HMM），通过维特比算法在状态路径中寻找后验概率最大的路径将声学特征映射为音素串82% similarVerifiedSpeech recognition technology has evolved from hidden Markov models (HMMs) to recurrent neural networks (RNNs) and then to end-to-end models.72% similarUnverifiedIn multimodal on-device inference, audio signals need to be converted into feature vectors via methods such as mel spectrograms, while motion data comes from the phone's accelerometer and gyroscope.71% similarVerifiedEnd-to-end voice models take audio directly as input and output, performing understanding and generation at the audio token level while preserving acoustic features such as intonation, speaking rate, sighs, and laughter.70% similarUnverifiedMicrophone arrays use beamforming technology to calculate the direction of sound by leveraging the time difference of arrival (TDOA) of the same sound source received by each microphone, thereby focusing on the sound from the target direction and suppressing noise.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/579671API
curl https://kongchang.com/api/v1/knowledge/claims/579671MCP
get_claim(id=579671)