| CPC G10L 15/063 (2013.01) [G10L 13/00 (2013.01); G10L 15/16 (2013.01)] | 19 Claims |
|
1. An operating method of an electronic device, the operating method comprising:
based on an utterance input of a specific speaker, obtaining target voice data;
determining a pre-trained model;
based on a preset criterion, determining at least one of an initial learning rate and a learning rate decay method of the target voice data;
based on the initial learning rate and the learning rate decay method, generating a target model by training the pre-trained model using the target voice data as training data, the pre-trained model being pre-trained to convert text into one or more audio signals;
generating output data by converting input text into an audio signal using the target model; and
outputting the output data.
|