Publication

判定支援システム、判定支援方法及びコンピュータープログラム (Applied in 02/05/2023)(特願2023-076257, 2023/May/2)

The patent is to enable more correct identification of the type of developmental disability. As a means of solution, a learning process is performed using writing information that indicates information about writing when a subject writes with an electronic pen, and correct answer information that indicates the type of developmental disability of the subject. The learned model obtained by this process is stored in a memory device, and the subject writes with an electronic pen. The estimation support can be equipped with an estimation unit that acquires writing information indicating information about the writing and estimates the type of developmental disability of the subject using the writing information and the learned model.

ユーザ認証プログラム、情報処理装置及びユーザ認証方法 (Applied in 11 /2022)(特願2022-187467)

1. Our dynamic gesture pattern is easy to remember. Since we are considering both the physiological and behavioral information of a person, it is not possible to create fake gesture of a person. So, It is reliable (low risk of leaking information).
2. Easy to remember and needs almost no special training.
3. We can install our system easily in any place (table, wall, etc). It requires a very small place.
4. It is a touchless system. So, it is very helpful to prevent infectious diseases like COVID-19.
5. All parts of the hand in our gesture pattern are easily visible to the leap motion. For this reason, we are getting very good features that are very necessary to the classification methods.
6. We can create various lengths of passwords. We can also use any permutation of the gesture pattern.
7. This gesture pattern is simple but has a strong discriminative capability.

(Accepted At 17/Jan/2024) 分類プログラム、分類装置及び分類方法特許No. 7421603, (特願2020-028085)

Parkinson’s disease is a movement disorder. It affects the nervous system, and symptoms become worse over time. If an individual can be diagnosed at an early stage in the development of Parkinson's disease, the treatment is more likely to be effective. In this context, we focus on the differential diagnosis of PD based on the handwriting. Handwriting consists of continuous, discontinuous, and Japanese characters depending on one or more strokes, i.e. tracing of the Archimedes spiral and sign curves, three-circle imitation, three-line marking, and free character writing. We have explored new features and investigated the discriminated between PD patient and healthy individuals using machine learning techniques. As the contribution of patent, in addition to the Spiral, Continuous tasks, Discontinuous tasks, and Katakana (character) tasks are compared. We use the following feature; (1) Time Series Features (kinematic feature) such as Writing Velocity, Acceleration, and Jerk, Pressure 1st and 2nd derivative, Azimuth 1st and 2nd derivative, Altitude 1st and 2nd derivative, Angle in Stroke (around 1 / 3 / 5 mm) (2) We use statistical representative value, such as Average, Standard Deviation, (3) We use DP Matching for each feature. We achieved around 95% recognition rate.

Parkinson's disease (PD) is generally taken into consideration as a disease that involves the movement, it can also be accompanied by motor symptoms, such as slowness of movement, tremors, and stiffness, and non-motor symptoms like depression, sleep problems and loss of smell. However, if an individual can be diagnosed at an early stage in the development of Parkinson's disease, the treatment is more likely to be effective. In this context, this paper demonstrates the different kinematics of handwriting, angle of stroke and dynamic programming (DP) features that can be used for differential diagnosis of PD. The database contains handwritten records of 19 patients and 17 healthy individuals who performed six different tasks. The tasks consisted of continuous, discontinuous, and Japanese characters depending on one or more strokes. These include tracing of the Archimedes spiral and sign curves, three-circle imitation, three-line marking, and free character writing. As for feature extraction, kinematic features are extracted from dynamic handwriting, we have explored new features based on the properties obtained by the first and second derivatives of the pen angle, the angle of stroke and the DP matching. Therefore, important features were investigated by the t-test and discriminated between PD patient and healthy individuals by comparing LSTM and SVM classifications.

(Accepted in 20th March 2015) (特許Japan No. 2011-018668) (特許Korea No. 10-2011-0113797)

Abstract: Handwriting Synthesis is one of the techniques of the cursive of the personal character creating artificially. Our new patent is the updated technology which creates the cursive, hand char. of the specific person (a few number) and the corresponding person can understanding his own writing and various style. This technique has the (1) “Individuality” which means the similarity between real handwriting and artificial generated character, (2) “Variability” which means the variety of the generated characters.