Speech synthesizing dolls for mimicking voices of parents and guardians of children
AssigneeSony Corporation
InventorsBrant Candelore, Mahyar Nejat
Life of this patent
Estimating the term…
Abstract
Speech synthesizing devices and methods are disclosed for using a doll or toy to mimic the voice of a parent or guardian of a child as reading a children's book. A speech synthesizer that employs an artificial neural network can be used to do so, with the artificial neural network being trained using audio recordings of the parent or guardian speaking as well as text corresponding to the words that are spoken by the parent or guardian in the audio recordings. The artificial neural network may then be used to produce various audio outputs in the voice of the parent or guardian as reading a children's book to the parent or guardian's child.
Claims15
A doll, comprising:
- at least one processor;
- at least one element on the doll and manipulable by hand to assume a first configuration to configure the doll to produce representations of a first person as speaking audio, the at least one element being manipulable to produce representations of a second person as speaking audio;
- and
- at least one speaker accessible to the at least one processor;
- the processor being programmed with instructions to:
- produce a representation of the first person as speaking audio responsive to the at least one element being in the first configuration and produce a representation of the second person as speaking audio responsive to the at least one element being in the second configuration, a microphone being accessible to the at least one processor, and wherein the instructions are executable to instruct the first and second persons to speak phrases provided by the doll to train respective deep neural networks (DNNs) for the respective first and second persons, the phrases as spoken by the first and second persons being detected by the microphone.
A doll comprising:
- at least one processor;
- a microphone accessible to the at least one processor;
- and
- at least one speaker accessible to the at least one processor;
- the processor being programmed with instructions to:
- produce a representation of at least a first person as speaking audio, the doll further comprising:
- a first hardware element with which the first person is associable such that upon association the first hardware element is manipulable to configure the doll to produce representations of the first person as speaking audio;
- and
- a second hardware element with which a second person is associable such that upon association the second hardware element is manipulable to configure the doll to produce representations of the second person as speaking audio, the second person being different from the first person, and wherein the instructions are executable to prompt to speak phrases provided by the doll to train respective deep neural networks (DNNs) for the respective first and second persons, the phrases as spoken by the first and second persons being detected by the microphone.
An apparatus, comprising:
- at least one computer readable storage medium that is not a transitory signal, the at least one computer readable storage medium comprising instructions executable by at least one processor to:
- present at least one perceptible prompt for first and second persons to speak phrases provided by a doll to train respective deep neural networks (DNNs) for the respective first and second persons, the phrases as spoken by the first and second persons being detected by a microphone;
- produce a representation of the first person as speaking audio according to training of at least one of the DNNs responsive to establishing a first configuration of the doll;
- and
- produce a representation of a second person as speaking audio according to training of at least one of the DNNs responsive to establishing a second configuration of the doll.
Dependencies are inferred from the claim text.