In:
Journal of the Acoustical Society of America, Acoustical Society of America (ASA), Vol. 140, No. 4_Supplement ( 2016-10-01), p. 3374-3375
Abstract:
Hand writing sound is one of noises existing in the real environment. Generally speaking, noises are meaningless; however, some noises, e.g., footsteps and a road noise, are effective to recognize circumstances. Hand writing sound also has some information to recognize a writing character. Previous study addressed the recognition of handwritten numeric characters by using writing sound, and the average recognition rate was 88.4%. We advanced previous research and examined the possibility of writer recognition from writing sound. We asked our subjects to write three target names. When subjects wrote on paper by a ballpoint pen, writing sounds were recorded. Three subjects chosen as the target name also wrote their own names under the different conditions, e.g., writing slowly and writing in another day. In our experiments, writer recognition was performed by the hidden Markov model. Acoustic features were mel-frequency cepstrum coefficients (MFCC), delta coefficients, and delta-delta (acceleration) coefficients. Experimental results showed that the writer recognition is possible; however, problems still exist such as how to deal with the difference between the writing conditions. In future works, it is necessary to increase learning data and to improve the recognition method.
Type of Medium:
Online Resource
ISSN:
0001-4966
,
1520-8524
Language:
English
Publisher:
Acoustical Society of America (ASA)
Publication Date:
2016
detail.hit.zdb_id:
1461063-2
Permalink