Methods for Encoding and Decoding Audio Signal Using Neural Network Model, and Devices for Performing the Same

摘要

Disclosed are a method for encoding and decoding an audio signal using a neural network model, a training method for a neural network model, and an encoder and decoder for performing the same. According to an embodiment of the present invention, an encoding method includes generating first characteristic information about the input signal using a cyclic encoding model; quantizing the first feature information to output quantized first feature information and a first bitstream; generating a first output signal from the quantized first feature information using a cyclic decoding model; generating second characteristic information for the input signal by using an acyclic encoding model; quantizing the second characteristic information to output quantized second characteristic information and a second bitstream; generating a second output signal from the quantized second feature information using an acyclic decoding model; determining an encoding mode using the input signal, the first and second output signals, and the first and second bitstreams; and outputting the entire bitstream by multiplexing the encoding mode bits with the first or second bitstream selected according to the encoding mode.