Singing assisting system, singing assisting method, and non-transitory computer-readable medium comprising instructions for executing the same

A non-transitory computer-readable medium, comprising a plurality of instructions, wherein when at least one processor of a computer system executes the instructions, the computer system executes steps of:

• detecting if a performed singing track appears in an ought-to-be-performed period, if not, executing a singing-continuing procedure, wherein the singing-continuing procedure is executed by being accompanied with an acoustics model comprising an encoder and a decoder, the acoustics model is trained by original-singer training data to obtain an original-singer acoustics model, and the acoustics model is trained by user training data to obtain a user acoustics model, the singing-continuing procedure comprises:
• converting an original-singer sound segment into an original-singer acoustics feature;
• inputting the original-singer acoustics feature to the encoder of the original-singer acoustics model;
• receiving an output of the encoder of the original-singer acoustics model by the decoder of the user acoustics model;
• obtaining a first user acoustics feature from an output of the decoder of the user acoustics model;
• and
• converting the first user acoustics feature into a synthesized singing track by a vocoder.

A singing assisting system, comprising:

• a sound receiving device, adapted to receive a performed singing track of a user;
• a processing device, adapted to detect if the performed singing track appears in an ought-to-be-performed period, if not, the processing device executing a singing-continuing procedure, wherein the singing-continuing procedure is executed by being accompanied with an acoustics model comprising an encoder and a decoder, the acoustics model is trained by original-singer training data to obtain an original-singer acoustics model, and the acoustics model is trained by user training data to obtain a user acoustics model, the singing-continuing procedure comprises:
• converting an original-singer sound segment into an original-singer acoustics feature;
• inputting the original-singer acoustics feature to the encoder of the original-singer acoustics model;
• receiving an output of the encoder of the original-singer acoustics model by the decoder of the user acoustics model;
• obtaining a first user acoustics feature from an output of the decoder of the user acoustics model;
• and
• converting the first user acoustics feature into a synthesized singing track by a vocoder;
• and
• a player device adapted to output the synthesized singing track.

A singing assisting method, comprising:

• receiving a performed singing track by a sound receiving device;
• determining if a performed pitch of the performed singing track is in consistent with an original-singer pitch of an original-singer singing track, if not, executing a pitch adjustment procedure to the performed singing track, wherein the pitch adjustment procedure is executed by being accompanied with a pitch model comprising a encoder and a decoder, the pitch model is trained by the original-singer training data to obtain an original-singer pitch model, and the pitch model is trained by the user training data to obtain a user pitch model, the pitch adjustment procedure comprises:
• converting the performed singing track into a user sound spectrum;
• converting the original-singer singing track into an original-singer sound spectrum;
• inputting the user sound spectrum to the encoder of the user pitch model to obtain a user acoustics feature;
• inputting the original-singer sound spectrum to the encoder of the original-singer pitch model to obtain an original-singer fundamental frequency;
• inputting the user acoustics feature and the original-singer fundamental frequency to the decoder of the user pitch model;
• obtaining an adjusted user sound spectrum from an output of the decoder of the user pitch model;
• and
• converting the adjusted user sound spectrum into a pitch-adjusted performed singing track;
• and
• outputting the synthesized singing track by a player device.

A non-transitory computer-readable medium, comprising a plurality of instructions, wherein when at least one processor of a computer system executes the instructions, the computer system executes steps of:

• determining if a performed pitch of the performed singing track is in consistent with an original-singer pitch of an original-singer singing track, if not, executing a pitch adjustment procedure to the performed singing track, wherein the pitch adjustment procedure is executed by being accompanied with a pitch model comprising a encoder and a decoder, the pitch model is trained by the original-singer training data to obtain an original-singer pitch model, and the pitch model is trained by the user training data to obtain a user pitch model, the pitch adjustment procedure comprises:
• converting the performed singing track into a user sound spectrum;
• converting the original-singer singing track into an original-singer sound spectrum;
• inputting the user sound spectrum to the encoder of the user pitch model to obtain a user acoustics feature;
• inputting the original-singer sound spectrum to the encoder of the original-singer pitch model to obtain an original-singer fundamental frequency;
• inputting the user acoustics feature and the original-singer fundamental frequency to the decoder of the user pitch model;
• obtaining an adjusted user sound spectrum from an output of the decoder of the user pitch model;
• and
• converting the adjusted user sound spectrum into a pitch-adjusted performed singing track.

A singing assisting system, comprising:

• a sound receiving device, adapted to receive a performed singing track of a user;
• a processing device, adapted to determine if a performed pitch of the performed singing track is in consistent with an original-singer pitch of an original-singer singing track, if not, the processing device executes a pitch adjustment procedure to the performed singing track, wherein the pitch adjustment procedure is executed by being accompanied with a pitch model comprising a encoder and a decoder, the pitch model is trained by the original-singer training data to obtain an original-singer pitch model, and the pitch model is trained by the user training data to obtain a user pitch model, the pitch adjustment procedure comprises:
• converting the performed singing track into a user sound spectrum;
• converting the original-singer singing track into an original-singer sound spectrum;
• inputting the user sound spectrum to the encoder of the user pitch model to obtain a user acoustics feature;
• inputting the original-singer sound spectrum to the encoder of the original-singer pitch model to obtain an original-singer fundamental frequency;
• inputting the user acoustics feature and the original-singer fundamental frequency to the decoder of the user pitch model;
• obtaining an adjusted user sound spectrum from an output of the decoder of the user pitch model;
• and
• converting the adjusted user sound spectrum into a pitch-adjusted performed singing track;
• and
• a player device adapted to output the pitch-adjusted performed singing track.