Reshape Dimensions Network for Speaker Recognition

TL;DR

ReDimNet is a novel neural network architecture that enhances speaker recognition by combining 1D and 2D feature processing through dimensionality reshaping, achieving state-of-the-art results with lower computational costs.

Contribution

The paper introduces ReDimNet, a scalable neural network architecture that uniquely integrates 1D and 2D feature maps via dimensionality reshaping for improved speaker recognition.

Findings

Achieves state-of-the-art speaker recognition performance.

Reduces computational complexity compared to existing models.

Offers scalable model sizes from 1 to 15 million parameters.

Abstract

In this paper, we present Reshape Dimensions Network (ReDimNet), a novel neural network architecture for extracting utterance-level speaker representations. Our approach leverages dimensionality reshaping of 2D feature maps to 1D signal representation and vice versa, enabling the joint usage of 1D and 2D blocks. We propose an original network topology that preserves the volume of channel-timestep-frequency outputs of 1D and 2D blocks, facilitating efficient residual feature maps aggregation. Moreover, ReDimNet is efficiently scalable, and we introduce a range of model sizes, varying from 1 to 15 M parameters and from 0.5 to 20 GMACs. Our experimental results demonstrate that ReDimNet achieves state-of-the-art performance in speaker recognition while reducing computational complexity and the number of model parameters.