Amazon reported mixed earnings but continues to dominate in the cloud and smart speaker space.
Inspired by JFA approach, Dehak et al. [7] propose a combination of speaker space and channel space.
In SAT training, acoustic models are trained on speaker-adapted training data, which are transformed into canonical speaker space by using speaker adaptation techniques, in this case, fMLLR.
The idea here is that Bose is finally ready to compete in the ultra-portable Bluetooth speaker space.
However, the number of mixture weights in GMMs for speaker identification is usually very large because each GMM should span not only speaker space but also phonetic space.
It trains acoustic models in a canonical speaker space based on the MLLR framework in order to obtain better feature transformation in the adaptation stage.
This paper presents a new approach to online adaptation of continuous density hidden Markov model (CDHMM) with a small amount of adaptation data based on speaker space model (SSM) evolution.
Both Amazon and Google have been operating in the smart speaker space for some time with the Echo and Home (respectively).
As the smart speaker space, and subsequent skill creation one, continue to heat up, the creation of your perfectly customized new smart speaker family member may be closer than you think.
