TY - GEN
T1 - What Did You Mean? An Evaluation of Online Character Recognition Approaches
AU - Koellner, Christopher
AU - Kurz, Marc
AU - Sonnleitner, Erik
N1 - Publisher Copyright:
© 2019 IEEE.
PY - 2019/10
Y1 - 2019/10
N2 - This paper presents a project that deals with online character recognition based on various time series such as acceleration, gyro and force. One important aspect of this work is the comparison and evaluation between a complex neural network-more precisely a Long-Short-Term-Memory network (LSTM)-and more comprehensible models such as K-Nearest Neighbours (KNN), Linear Discriminant Analysis (LDA) and NaiveBayes (NB). LSTMs have shown to not only be suitable for speech recognition but also for the recognition of handwriting based on time series due to its ability to learn long-Time dependencies of the provided data. Online handwriting describes the usage of a digitizer in order to convert analogue sensor data to a digital representation of the handwriting. The digitizer used in our approach refers to a sensor enhanced ball pen developed by STA-BILO. The data set used for training consists of approximately 20,000 lowercase English alphabet characters in total contributed by 15 different people. The discussed approaches are evaluated and compared against each other using the recorded dataset, while taking recognition rates (i.e., accuracy) into account.
AB - This paper presents a project that deals with online character recognition based on various time series such as acceleration, gyro and force. One important aspect of this work is the comparison and evaluation between a complex neural network-more precisely a Long-Short-Term-Memory network (LSTM)-and more comprehensible models such as K-Nearest Neighbours (KNN), Linear Discriminant Analysis (LDA) and NaiveBayes (NB). LSTMs have shown to not only be suitable for speech recognition but also for the recognition of handwriting based on time series due to its ability to learn long-Time dependencies of the provided data. Online handwriting describes the usage of a digitizer in order to convert analogue sensor data to a digital representation of the handwriting. The digitizer used in our approach refers to a sensor enhanced ball pen developed by STA-BILO. The data set used for training consists of approximately 20,000 lowercase English alphabet characters in total contributed by 15 different people. The discussed approaches are evaluated and compared against each other using the recorded dataset, while taking recognition rates (i.e., accuracy) into account.
KW - Long Short Term Memory Network
KW - Machine Learning
KW - Mobile Platform Evaluation
KW - Online Handwriting Recognition
UR - http://www.scopus.com/inward/record.url?scp=85077580341&partnerID=8YFLogxK
U2 - 10.1109/WiMOB.2019.8923384
DO - 10.1109/WiMOB.2019.8923384
M3 - Conference contribution
T3 - International Conference on Wireless and Mobile Computing, Networking and Communications
BT - 2019 International Conference on Wireless and Mobile Computing, Networking and Communications, WiMob 2019
PB - IEEE Computer Society
T2 - 15th International Conference on Wireless and Mobile Computing, Networking and Communications, WiMob 2019
Y2 - 21 October 2019 through 23 October 2019
ER -