docs(touch_digit): add touch principle

Author: lijunru-hub

docs/_static/ai/touch_kit_pad.png

@@ -9,6 +9,35 @@ Touch Principle and Data Acquisition
 Touch Principle
 ^^^^^^^^^^^^^^^^^^
 
+The `ESP_Touch_Kit_Touchpad <https://dl.espressif.com/dl/schematics/SCH_ESP-Touch-Kit-Touchpad_V1.0_20210406.pdf>`_ is used as the touch pad. This touchpad consists of a 6*7 array of touch channels.
+
+.. figure:: ../../_static/ai/touch_kit_pad.png
+    :align: center
+
+    Physical Touch Pad
+
+When a finger moves across the touch pad, it changes the capacitance values of the touch channels, allowing us to detect the finger position by monitoring these capacitance changes.
+
+Detection Algorithm
+^^^^^^^^^^^^^^^^^^^^
+
+1. Due to hardware variations, each channel has different maximum and minimum trigger values. Therefore, the capacitance values need to be normalized. This is done by recording the maximum and minimum values as the finger slides across the pad.
+
+2. During finger movement, one channel will show the highest rate of capacitance change. By identifying this channel and its adjacent channel with the next highest change rate, we can locate two neighboring channels.
+
+3. Using a ratio-based calculation between these two channels' values, we can determine the relative coordinate of the finger along that direction (the offset from the center point between the two channels).
+
+.. math::
+
+   x = \frac{Fa - Fb}{Fa + Fb}
+
+4. By applying the above steps, we can obtain relative coordinates in both directions, thus determining the finger's position.
+
+.. note::
+
+    An appropriate trigger threshold needs to be set to determine whether the finger is moving on the touch pad.
+
+5. When finger lift-off is detected, the drawn data can be saved.
 
 Data Collection
 ^^^^^^^^^^^^^^^^^^^
@@ -284,7 +313,7 @@ To facilitate model debugging, ESP-DL provides the functionality to add test dat
 .. code-block:: bash
 
     test outputs value:
-    %23, shape: [1, 10], exponents: [0], 
+    %23, shape: [1, 10], exponents: [0],
     value: array([9.85415445e-34, 1.92874989e-22, 7.46892081e-43, 1.60381094e-28,
         3.22134028e-27, 1.05306175e-20, 4.07960022e-41, 1.42516404e-21,
         2.38026637e-26, 1.00000000e+00, 0.00000000e+00, 0.00000000e+00],

docs/en/ai/touch_digit_recognition.rst

@@ -9,6 +9,35 @@ Touch 原理与数据采集
 Touch 原理
 ^^^^^^^^^^^^^
 
+采用 `ESP_Touch_Kit_Touchpad <https://dl.espressif.com/dl/schematics/SCH_ESP-Touch-Kit-Touchpad_V1.0_20210406.pdf>`_ 作为 Touch 触摸板。该触摸板有 6*7 的 Touch 通道组成。
+
+.. figure:: ../../_static/ai/touch_kit_pad.png
+    :align: center
+
+    Touch 触控板实物图
+
+当手指在 Touch 触摸板上移动时，会改变 Touch 通道的电容值，这时候我们就可以通过检测电容值的变化来判断手指的位置。
+
+检测算法
+^^^^^^^^^^^^
+
+1. 因为硬件差异，各个通道触发的最大值和最小值不同，所以需要对各个通道的电容值进行归一化处理。通过手指在上面滑动，来记录各个通道的最大值和最小值。
+
+2. 在手指滑动的过程中，会有一路通道的电容值变化率最大，这时候再取相邻变化率最大的通道，可以获取到两条相邻的通道。
+
+3. 将两个通道的值进行差比和运算，可以获取到手指在该方向上的相对坐标（较两通道中心点的偏移量）。
+
+.. math::
+
+   x = \frac{Fa - Fb}{Fa + Fb}
+
+4. 通过上述步骤，可以获取到手指在两个方向上的相对坐标，从而可以确定手指的位置。
+
+.. note::
+
+    需要设定合适的触发阈值，来判断手指是否在 Touch 触摸板上移动。
+
+5. 当检测到手指抬起，即可将绘制的数据进行保存。
 
 数据采集
 ^^^^^^^^^^^^
@@ -284,7 +313,7 @@ ESP-PPQ 环境配置
 .. code-block:: bash
 
     test outputs value:
-    %23, shape: [1, 10], exponents: [0], 
+    %23, shape: [1, 10], exponents: [0],
     value: array([9.85415445e-34, 1.92874989e-22, 7.46892081e-43, 1.60381094e-28,
         3.22134028e-27, 1.05306175e-20, 4.07960022e-41, 1.42516404e-21,
         2.38026637e-26, 1.00000000e+00, 0.00000000e+00, 0.00000000e+00],

docs/zh_CN/ai/touch_digit_recognition.rst

@@ -33,8 +33,8 @@ extern "C" void app_main(void)
         ret = nvs_flash_init();
     }
     ESP_ERROR_CHECK(ret);
-    /* Initialize button */
 
+    /* Initialize button */
     const button_config_t btn_cfg = {};
     const button_gpio_config_t btn_gpio_cfg = {
         .gpio_num = 0,

examples/ai/esp_dl/touchpad_digit_recognition/main/main.cpp

@@ -26,10 +26,8 @@
 
 static const char *TAG = "touch_digit";
 
+#define REALLY_DATA_PRINT 0  // If you need to collect data using the host computer, please set this macro to true
 #define SCAN_TIME_MEASURE 0
-#define REALLY_DATA_PRINT 0
-#define TOUCH_DIGIT_DATA_PRINT 1
-#define NORMALIZE_DATA_PRINT 1
 
 #define CHANNEL_NUM 13
 #define CHANNEL_LIST {1,2,3,4,5,6,7,8,9,10,11,12,13}
@@ -213,7 +211,7 @@ static void touch_digit_task(void *arg)
                 }
             }
             printf("\n");
-#endif
+#else
 
 #if SCAN_TIME_MEASURE
             uint32_t now_time = esp_timer_get_time();
@@ -247,7 +245,6 @@ static void touch_digit_task(void *arg)
                 break;
             case END_WRITE:
                 // send data to dl inference
-
                 g_image.print();
                 image_data_t image_data;
                 image_data.size = g_image.col_length * g_image.row_length;
@@ -263,6 +260,7 @@ static void touch_digit_task(void *arg)
             default:
                 break;
             }
+#endif
         }
     }
 }
@@ -287,7 +285,7 @@ void touch_digit_recognition_task(void *arg)
 
     while (1) {
         if (xQueueReceive(xImageQueue, &image_data, portMAX_DELAY) == pdTRUE) {
-            // printf("Result:%d\n", touch_digit_recognition->predict(image_data.data));
+            printf("Result:%d\n", touch_digit_recognition->predict(image_data.data));
             digital_tube_write_num(0, touch_digit_recognition->predict(image_data.data));
             delete [] image_data.data;
         }