Channel Shuffle Neural Architecture Search for Key Word Spotting

Proceedings Paper Computer Science, Artificial Intelligence

EFFICIENT DYNAMIC FILTER FOR ROBUST AND LOW COMPUTATIONAL FEATURE EXTRACTION

Donghyeon Kim et al.

Summary: In this paper, we propose an efficient dynamic filter by using chunking and separable convolutions to enhance its performance. We also propose a dynamic filter based attention pooling method. Experimental results demonstrate that our proposed methods outperform the recently developed models in unseen environments.

2022 IEEE SPOKEN LANGUAGE TECHNOLOGY WORKSHOP, SLT (2022)

Add to Collection

Proceedings Paper Audiology & Speech-Language Pathology

Broadcasted Residual Learning for Efficient Keyword Spotting

Byeonggeun Kim et al.

Summary: Keyword spotting is an important research field, and the proposed broadcasted residual learning method achieves high accuracy with small model size and computational load.

INTERSPEECH 2021 (2021)

Add to Collection

Proceedings Paper Computer Science, Artificial Intelligence

EFFICIENT KEYWORD SPOTTING BY CAPTURING LONG-RANGE INTERACTIONS WITH TEMPORAL LAMBDA NETWORKS

Biel Tura et al.

Summary: This study explores the application of Lambda networks in keyword spotting, proposing a novel ResNet-based model with temporal Lambda layers and uni-dimensional temporal convolutions. The model achieves state-of-the-art accuracies on the Google Speech Commands dataset and is significantly lighter and faster than Transformer and convolutional models. This is the first attempt to utilize the Lambda framework in the speech domain, opening up possibilities for further research in new interfaces based on this architecture.

2021 IEEE AUTOMATIC SPEECH RECOGNITION AND UNDERSTANDING WORKSHOP (ASRU) (2021)

Add to Collection

Proceedings Paper Computer Science, Artificial Intelligence

FP-NAS: Fast Probabilistic Neural Architecture Search

Zhicheng Yan et al.

Summary: Researchers proposed a method called Fast Probabilistic NAS (FP-NAS) which allows more efficient model search in multi-variate space with fewer samples and faster speed, while achieving better model performance compared to other methods.

2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021 (2021)

Add to Collection

Proceedings Paper Acoustics

AUTOKWS: KEYWORD SPOTTING WITH DIFFERENTIABLE ARCHITECTURE SEARCH

Bo Zhang et al.

Summary: This paper proposes a method using neural architecture search to discover more efficient networks, which can effectively reduce power consumption of smart audio devices.

2021 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP 2021) (2021)

Add to Collection

Proceedings Paper Acoustics