20.500.12556/DiRROS-18556 Hardware–software co-design of an audio feature extraction pipeline for machine learning applications Keyword spotting is an important part of modern speech recognition pipelines. Typical contemporary keyword-spotting systems are based on Mel-Frequency Cepstral Coefficient (MFCC) audio features, which are relatively complex to compute. Considering the always-on nature of many keyword-spotting systems, it is prudent to optimize this part of the detection pipeline. We explore the simplifications of the MFCC audio features and derive a simplified version that can be more easily used in embedded applications. Additionally, we implement a hardware generator that generates an appropriate hardware pipeline for the simplified audio feature extraction. Using Chisel4ml framework, we integrate hardware generators into Python-based Keras framework, which facilitates the training process of the machine learning models using our simplified audio features. FPGA MFCC keyword spotting chisel true false true MDPI Švica Angleški jezik Ni določen © 2024 by the authors. Neznano 2024-03-25 10:55:57 2024-03-25 10:55:58 2024-03-26 03:35:54 0000-00-00 00:00:00 2024 0 0 Nasl. z nasl. zaslona; Opis vira z dne 26. 2. 2024; str. 1-14 5 13 24. Feb. 2024 0000-00-00 Zaloznikova Objavljeno NiDoloceno 2024-01-31 2024-02-22 2024-02-24 004 2079-9292 10.3390/electronics13050875 186803203 https://www.mdpi.com/2079-9292/13/5/875 1 e20e97b0-ea8d-11ee-9dd4-001a4af901a5 https://dirros.openscience.si/Dokument.php?lang=slv&id=25282 RAZ_Vreca_Jure_2024.pdf RAZ_Vreca_Jure_2024.pdf 1 F09ED506C9C50118B6B65AC40E9F9684 f28dba42c5542866ba17d1681f0cf3004fe588df87144c579d081e40e815cfc2 c1feaf8d-ea8f-11ee-9dd4-001a4af901a5 https://dirros.openscience.si/Dokument.php?lang=slv&id=25283 Institut Jožef Stefan 0 0 0