Scholarlink

<p>Flexible piezoresistive tactile sensors hold broad application prospects in the fields of wearable sensing and human–computer interaction due to their high sensitivity, simple structure, and excellent mechanical flexibility. This paper presents a high-performance piezoresistive sensor mainly consists of a conductive composite made by depositing multi-walled carbon nanotube (MWCNT) and carbon black (CB) to the melamine sponge (MS), where the MS worked as the flexible substrate and the MWCNT and CB used as the conductive filler. In the experment, the sensor exhibits great ability with high sensitivity (4.767%·kPa⁻<sup>1</sup> in the 0–10&#xa0;kPa range), fast response/recovery time (240&#xa0;ms/90&#xa0;ms), low hysteresis (9.39%), and fairly good cycling stability (2000 cycles) within the range of 0–100&#xa0;kPa. This work further integrated four sensing units into a flexible wristband to develop a wearable wrist sensor. A dataset containing 5200 samples was constructed by collecting four-channel temporal voltage signals generated during in-air writing of 26 letters (A–Z). To effectively extract spatiotemporal features from the multichannel signals, a hybrid CNN-LSTM neural network model was proposed, wherein the CNN layers are responsible for extracting spatial correlation features among channels, and the LSTM layers capture long-term temporal dependencies of the gesture motions. Experimental results show that the model achieves an accuracy of 98.65% in the letter classification task, validating the effectiveness and potential of the proposed wrist sensor and recognition algorithm for in-air handwriting recognition applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

In-air handwriting recognition based on the MWCNT/CB wrist sensor with a CNN-LSTM model

  • 期刊论文

  • 发表日期：

    2026年4月10日

查看全文
  • Yang Song,
  • Tengsheng Zhang,
  • Feilu Wang,
  • Changming Huang,
  • Wenjun He,
  • Yangfan Chen

摘要

Flexible piezoresistive tactile sensors hold broad application prospects in the fields of wearable sensing and human–computer interaction due to their high sensitivity, simple structure, and excellent mechanical flexibility. This paper presents a high-performance piezoresistive sensor mainly consists of a conductive composite made by depositing multi-walled carbon nanotube (MWCNT) and carbon black (CB) to the melamine sponge (MS), where the MS worked as the flexible substrate and the MWCNT and CB used as the conductive filler. In the experment, the sensor exhibits great ability with high sensitivity (4.767%·kPa⁻1 in the 0–10 kPa range), fast response/recovery time (240 ms/90 ms), low hysteresis (9.39%), and fairly good cycling stability (2000 cycles) within the range of 0–100 kPa. This work further integrated four sensing units into a flexible wristband to develop a wearable wrist sensor. A dataset containing 5200 samples was constructed by collecting four-channel temporal voltage signals generated during in-air writing of 26 letters (A–Z). To effectively extract spatiotemporal features from the multichannel signals, a hybrid CNN-LSTM neural network model was proposed, wherein the CNN layers are responsible for extracting spatial correlation features among channels, and the LSTM layers capture long-term temporal dependencies of the gesture motions. Experimental results show that the model achieves an accuracy of 98.65% in the letter classification task, validating the effectiveness and potential of the proposed wrist sensor and recognition algorithm for in-air handwriting recognition applications.