Continuous Motion Numeral Recognition Using RNN Architecture in Air-Writing Environment

Rahman, Adil; Roy, Prasun; Pal, Umapada

doi:10.1007/978-3-030-41404-7_6

Cited by 5 publications

(10 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The current TS-D model outperforms our previous results [22] of single and multi-digit English numeral recognition tests by 0.30% and 2.38%, respectively. We also Average accuracy = ∑ (SD × SA) TS × 100%, observe a significant increase of 26.05% and 26.79% in the single and multi digit recognition tests of our baseline model (TS-A), respectively.…”

Section: Resultscontrasting

confidence: 51%

“…Chen et al in [8,9] adopt a window-based algorithm where the window properties and operations are based on empirically set time/velocity constants, both of which are user-dependent factors. In our previous paper [22], we proposed a sliding window-based algorithm to segment characters traced in a continuous motion. However, prediction was done using only sequential coordinate information and details like velocity and time were not taken into consideration owing to its variability among different users, making it a poor classification feature.…”

Section: Input Recognitionmentioning

confidence: 99%

“…In our previous study [22], we proposed a priority scheme which capitalized on the way digits are generally written. Using Pendigits and our custom air-writing dataset ISI-Air, both of which are online English numeral datasets containing sequential motion coordinate information of each digit, we create an English numeral Priority Matrix, as illustrated in Fig.…”

Section: Priority Schemementioning

confidence: 99%

“…While certain stray motions exist which add to the overall noise in the system, a majority of it can be attributed to the transition noise. To counter this issue, in our previous paper [22] we created a noise dataset consisting of every possible transition noise between any two English numerals using the same method as used in the creation of ISI-Air which was referred to as NZ-1 in [22]. Owing to the variability of the real-time transition noise, the previous dataset was not very effective.…”

Section: Noise Datasetmentioning

confidence: 99%

“…An earlier version of this paper was presented at the 5th Asian Conference on Pattern Recognition, 2019 [22], in which we proposed a system to recognize a string of English numerals written in a continuous motion in an air-writing environment using a sliding-window-based algorithm and RNN-LSTM (Recurrent Neural Network/ Long Short-Term Memory) network. Our proposed methodology was inexpensive owing to the fact that it utilised only a generic webcam as its source of input and no external sensors.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Air Writing: Recognizing Multi-Digit Numeral String Traced in Air Using RNN-LSTM Architecture

Rahman¹,

Roy

Pal

2021

SN COMPUT. SCI.

Self Cite

View full text Add to dashboard Cite

Air writing provides a more natural and immersive way of interacting with devices, with the potential of having significant application in fields like augmented reality and education. However, such systems often rely on expensive hardware, making them less accessible for general purposes. In this study, we propose a robust and inexpensive system for the recognition of multi-digit numerals traced in an air-writing environment which uses only a generic device camera for input. We employ a sliding window-based algorithm to isolate a small segment of the input for processing. A dual network configuration consisting of RNN-LSTM networks are used for noise elimination and digit recognition. We conduct our experiments on English numerals using the MNIST dataset as the baseline model to allow easy adaptability of our method. Our results are further improved by the use of Pendigits and ISI-Air online datasets. We observed a drop in accuracy with increase in the number of digits owing to the accumulation of transition noise. However, bi-directional scanning considerably reduces the impact of such noise on the recognition accuracy. Under standard conditions, our system produced an accuracy of 98.75% and 85.27% for single and multi-digit English numerals, respectively. Incorporation of selective frame skipping in the sliding window algorithm resulted in a 60% reduction in computational time, significantly improving the system performance. We provide a link to the source code of our system at the end of this paper.

show abstract

Section: Resultscontrasting

confidence: 51%