This paper proposes an anomaly detection (AD) algorithm that can discriminate stylus‐touch based on capacitive touch screen panel. The digital value acquired from an analog‐to‐digital converter (ADC) are transferred to an autoencoder including an encoder and a decoder. While the encoder classifies only two classes of a no‐touch and a finger‐touch, the decoder reconstructs the similar sequence to the input one according to the encoder's decision. Because the touch sequences caused by the stylus are not trained, the large difference between input and output sequences is used to discriminate the stylus‐touch from finger‐touch and no‐touch. The proposed method is evaluated by means of an 8‐inch capacitive touch panel, an AD touch detection board, and a stylus board. At the sequence length of 16 touch samples, the measured bit error rate (BER) of less than 10⁻⁶ for each touch case is equivalent to the previous support vector machine (SVM) scheme whereas the number of multipliers is dramatically reduced to 16, compared with 400 of the previous SVM method.

中文翻译：

---

基于异常检测算法的主动触控触摸识别方案

本文提出了一种基于电容式触摸屏面板的可区分手写笔触摸的异常检测算法。从模数转换器（ADC）获得的数字值被传输到包括编码器和解码器的自动编码器。编码器仅将无触摸和手指触摸分为两类，而解码器根据编码器的决定将与输入的相似的序列重建。由于触控笔引起的触摸顺序没有经过训练，因此输入和输出顺序之间的巨大差异可用来区分触控笔触摸与手指触控和非触控。通过8英寸电容式触摸面板，AD触摸检测板和手写板对所提出的方法进行评估。在16个触摸样本的序列长度上，每个触摸情况的^-6等效于以前的支持向量机（SVM）方案，而乘法器的数量则大大减少到16，而以前的SVM方法为400。