End-to-end Keyword Spotting using Xception-1d

TL;DR

This paper presents an end-to-end keyword spotting system using an adapted Xception-1D model, achieving state-of-the-art accuracy of 96% across 35 categories, surpassing human performance.

Contribution

The work adapts the Xception architecture for audio keyword spotting, demonstrating its effectiveness and achieving superior accuracy in complex classification tasks.

Findings

Achieved 96% accuracy on 35-category keyword classification

Outperformed human annotation in complex tasks

Validated the effectiveness of Xception-1D for audio analysis

Abstract

The field of conversational agents is growing fast and there is an increasing need for algorithms that enhance natural interaction. In this work we show how we achieved state of the art results in the Keyword Spotting field by adapting and tweaking the Xception algorithm, which achieved outstanding results in several computer vision tasks. We obtained about 96\% accuracy when classifying audio clips belonging to 35 different categories, beating human annotation at the most complex tasks proposed.