3 ms·
> Gaussian mixture models In what fields did neural networks replace Gaussian mixtures?
by ForceBru 10mo ago
> Gaussian mixture models
In what fields did neural networks replace Gaussian mixtures?
- brandonb 10mo agoThe acoustic model of a speech recognizer used to be a GMM, which mapped a pre-processed acoustic signal vector (generally MFCCs-Mel-Frequency Cepstral Coefficients) to an HMM state. Now those layers are neural nets, so acoustic pre-processing, GMM, and HMM are all subsumed by the neural network and trained end-to-end. One early piece of work here was DeepSpeech2 (2015): https://arxiv.org/pdf/1512.02595 https://arxiv.org/pdf/1512.02595
- ForceBru 10mo agoInteresting, thanks!