seminar problem

2
SEMINAR PROBLEM Sensor gesture from phone (use phone for gesture) Simple app: start --> gesture --> stop --> show gesture 1. Data source - Get sensor data from phone (phone: android - Android Studio IDE): accelerometer, magnetometer, gyroscope [angle speed] --> max frequency [max sample get per sec] (from max freq use filtering to round down freq with best result) o Make app to record sample data --> labeling o Decision with bin[1,0] 2. Data storage+format - Use csv (excel like) (google csv extension) - Google “blob”: block of binary (※SQL) - using database for auto - Convert text data to database 3. Fit data to learning engine (PYTHON) - 1 lib to communicate with database --> put in SVM learning --< get model o Get used to SVM lib of python o Convert data to suitable input for python lib - Cross validation to overcome overfitting (k-fold validation upgrade) 4. Demo engine - Python got TEST - PROBLEM: From python to java → ANDROID To do list 1. Imply to 3 2. Read FML --> SVM 3. Learn Python // Android app (sci-kit package) 4. Handwriting recognition 5. Schedule a. 1+2: 2 months b. 3: 3 months (learn = 2 + combine = 1+2+3)

Upload: nguyen-duc-hiep

Post on 17-Feb-2016

214 views

Category:

Documents


0 download

DESCRIPTION

Conver

TRANSCRIPT

Page 1: Seminar Problem

SEMINAR PROBLEM

Sensor gesture from phone (use phone for gesture)

Simple app: start --> gesture --> stop --> show gesture

1. Data source- Get sensor data from phone (phone: android - Android Studio IDE): accelerometer,

magnetometer, gyroscope [angle speed] --> max frequency [max sample get per sec] (from max freq use filtering to round down freq with best result)

o Make app to record sample data --> labelingo Decision with bin[1,0]

2. Data storage+format- Use csv (excel like) (google csv extension)- Google “blob”: block of binary ( SQL) - using database for auto※- Convert text data to database

3. Fit data to learning engine (PYTHON)- 1 lib to communicate with database --> put in SVM learning --< get model

o Get used to SVM lib of pythono Convert data to suitable input for python lib

- Cross validation to overcome overfitting (k-fold validation upgrade)4. Demo engine

- Python got TEST- PROBLEM: From python to java → ANDROID

To do list

1. Imply to 32. Read FML --> SVM3. Learn Python // Android app (sci-kit package)4. Handwriting recognition5. Schedule

a. 1+2: 2 monthsb. 3: 3 months (learn = 2 + combine = 1+2+3)

6. Code convention Android + Python (STD) ※7. Repo ※8. Skype (+12g) ※