smartwatch_gestures,smartwatch_gestures

  • 설명 :

SmartWatch Gestures Dataset 은 팔 제스처를 사용하여 모바일 애플리케이션과 상호 작용하기 위한 여러 제스처 인식 알고리즘을 평가하기 위해 수집되었습니다.

8명의 다른 사용자가 총 3200개의 시퀀스에 대해 20개의 서로 다른 제스처를 20번 반복했습니다. 각 시퀀스에는 1세대 Sony SmartWatch™의 3축 가속도계의 가속 데이터와 Android 기기에서 사용할 수 있는 다양한 클록 소스의 타임스탬프가 포함되어 있습니다. 스마트워치는 사용자의 오른쪽 손목에 착용했습니다. 제스처는 모든 반복의 시작과 끝에서 스마트 워치 화면을 탭하여 수행하는 사용자에 의해 수동으로 분할되었습니다.

나뉘다
'train' 3,251
  • 기능 구조 :
FeaturesDict({
    'attempt': uint8,
    'features': Sequence({
        'accel_x': float64,
        'accel_y': float64,
        'accel_z': float64,
        'time_event': uint64,
        'time_millis': uint64,
        'time_nanos': uint64,
    }),
    'gesture': ClassLabel(shape=(), dtype=int64, num_classes=20),
    'participant': uint8,
})
  • 기능 문서 :
특징 수업 모양 D타입 설명
풍모Dict
시도 텐서 uint8
특징 순서
기능/accel_x 텐서 float64
기능/accel_y 텐서 float64
기능/accel_z 텐서 float64
기능/time_event 텐서 uint64
기능/time_millis 텐서 uint64
기능/time_nanos 텐서 uint64
몸짓 클래스 레이블 int64
참가자 텐서 uint8
  • 인용 :
@INPROCEEDINGS{
  6952946,
  author={Costante, Gabriele and Porzi, Lorenzo and Lanz, Oswald and Valigi, Paolo and Ricci, Elisa},
  booktitle={2014 22nd European Signal Processing Conference (EUSIPCO)},
  title={Personalizing a smartwatch-based gesture interface with transfer learning},
  year={2014},
  volume={},
  number={},
  pages={2530-2534},
  doi={} }