Voice conversion learning device, voice conversion device, method, and program

A voice conversion system, the system comprising a processor configured to execute operations comprising:

• encoding comprising estimating a latent variable series from input sound feature value series in a conversion-source voice signal and an attribution code indicating attribution of the conversion-source voice signal, the encoding further comprises estimating a latent variable series from input of a sound feature value series and an attribution code;
• and
• decoding comprising reconfiguring the sound feature value series from input of the latent variable series and the attribution code, the encoding and the decoding being previously learned to maximize a value of an objective function, on the basis of:
• parallel data of a sound feature value series in a conversion-source voice signal and a latent variable series in the conversion-source voice signal, and
• an attribution code indicating attribution of the conversion-source voice signal,
• wherein the objective function being represented using:
• attribution code similarity of a sound feature value series reconfigured by the decoding from input of an error between the sound feature value series reconfigured by the decoding and the sound feature value series in the conversion-source voice signal in the parallel data,
• a distance between the latent variable series estimated by the encoding and the latent variable series in the conversion-source voice signal in the parallel data, and
• any attribution code, the attribution code similarity being similarity to the any attribution code identified by an attribution identifier.

A voice conversion method, the method comprising:

• learning:
• encoding, on the basis of:
• parallel data of a sound feature value series in a conversion-source voice signal and a latent variable series in the conversion-source voice signal, and
• an attribution code indicating attribution of the conversion-source voice signal, estimating a latent variable series from input of the sound feature value series and an attribution code, and
• decoding further comprising reconfiguring the sound feature value series from input of the latent variable series and the attribution code,
• wherein the learning includes the encoding and decoding to maximize a value of an objective function,
• wherein the objective function being represented using:
• attribution code similarity of the sound feature value series reconfigured by the decoding from input of an error between the sound feature value series reconfigured by the decoding and the sound feature value series in the conversion-source voice signal in the parallel data,
• a distance between the latent variable series estimated by the encoding and the latent variable series in the conversion-source voice signal in the parallel data, and
• any attribution code, the attribution code similarity being similarity to the any attribution code identified by an attribution identifier.

A computer-readable non-transitory recording medium storing a computer-executable program for a voice conversion learning system that when executed by a processor cause the computer-executable program to:

• learn:
• encoding, on the basis of:
• parallel data of a sound feature value series in a conversion-source voice signal and a latent variable series in the conversion-source voice signal, and
• an attribution code indicating attribution of the conversion-source voice signal,
• estimating a latent variable series from input of the sound feature value series and the attribution code, and
• decoding comprising reconfiguring the sound feature value series from input of the latent variable series and the attribution code,
• wherein the learning further comprises learning the encoding and the decoding to maximize a value of an objective function,
• wherein the objective function being represented using:
• attribution code similarity of the sound feature value series reconfigured by the decoding from input of an error between the sound feature value series reconfigured by the decoding and the sound feature value series in the conversion-source voice signal in the parallel data,
• a distance between the latent variable series estimated by the encoding and the latent variable series in the conversion-source voice signal in the parallel data, and
• any attribution code, the attribution code similarity being similarity to the any attribution code identified by an attribution identifier.