TY - GEN
T1 - Towards the on-device Handwriting Trajectory Reconstruction of the Sensor Enhanced Pen
AU - Serdyuk, Alexey
AU - Kreb, Fabian
AU - Hiegle, Micha
AU - Harbaum, Tanja
AU - Becker, Jurgen
AU - Imbert, Florent
AU - Soullard, Yann
AU - Tavenard, Romain
AU - Anquetil, Eric
AU - Barth, Jens
AU - Kampf, Peter
N1 - Publisher Copyright:
© 2023 IEEE.
PY - 2023/1/1
Y1 - 2023/1/1
N2 - Performing handwriting trajectory regression from inertial data using Deep Neural Network (DNN) on an embedded device is a very challenging task, since the network accuracy is prone to imperfections in the weights and needs a significant amount of parameters to be able to regress. In this work, we apply and compare different quantization techniques and Mitchell logarithmic multiplication approximation in order to enable the on-device inference. We show that it is possible to perform the inference of the TCN-based regression model using only 8-bit fixed-point quantization without significant reconstruction precision loss and that the accuracy degradation of the approximate multiplication can be partially compensated with Quantization-aware Training (QAT). Finally, we demonstrate that the compressed models can be integrated into an off-the-shelf commercial Systems-on-Chip with minimal use of FPU and requiring only 460 KB of the ROM size for the TCN-49 configuration.
AB - Performing handwriting trajectory regression from inertial data using Deep Neural Network (DNN) on an embedded device is a very challenging task, since the network accuracy is prone to imperfections in the weights and needs a significant amount of parameters to be able to regress. In this work, we apply and compare different quantization techniques and Mitchell logarithmic multiplication approximation in order to enable the on-device inference. We show that it is possible to perform the inference of the TCN-based regression model using only 8-bit fixed-point quantization without significant reconstruction precision loss and that the accuracy degradation of the approximate multiplication can be partially compensated with Quantization-aware Training (QAT). Finally, we demonstrate that the compressed models can be integrated into an off-the-shelf commercial Systems-on-Chip with minimal use of FPU and requiring only 460 KB of the ROM size for the TCN-49 configuration.
KW - Embedded Systems
KW - Handwriting Reconstruction
KW - Human Machine Interfaces
KW - Internet of Things
KW - Temporal Convolutional Network
UR - https://www.scopus.com/pages/publications/85195361925
U2 - 10.1109/WF-IoT58464.2023.10539488
DO - 10.1109/WF-IoT58464.2023.10539488
M3 - Conference contribution
AN - SCOPUS:85195361925
T3 - 2023 IEEE World Forum on Internet of Things: The Blue Planet: A Marriage of Sea and Space, WF-IoT 2023
BT - 2023 IEEE World Forum on Internet of Things
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 9th IEEE World Forum on Internet of Things, WF-IoT 2023
Y2 - 12 October 2023 through 27 October 2023
ER -