demo链接：https://download.csdn.net/download/meixi_android/10690974

工具类：

public class StepDetector implements SensorEventListener {//存放三轴数据float[] oriValues = new float[3];final int ValueNum = 4;//用于存放计算阈值的波峰波谷差值float[] tempValue = new float[ValueNum];int tempCount = 0;//是否上升的标志位boolean isDirectionUp = false;//持续上升次数int continueUpCount = 0;//上一点的持续上升的次数，为了记录波峰的上升次数int continueUpFormerCount = 0;//上一点的状态，上升还是下降boolean lastStatus = false;//波峰值float peakOfWave = 0;//波谷值float valleyOfWave = 0;//此次波峰的时间long timeOfThisPeak = 0;//上次波峰的时间long timeOfLastPeak = 0;//当前的时间long timeOfNow = 0;//当前传感器的值float gravityNew = 0;//上次传感器的值float gravityOld = 0;//动态阈值需要动态的数据，这个值用于这些动态数据的阈值final float InitialValue = (float) 1.3;//初始阈值float ThreadValue = (float) 2.0;//波峰波谷时间差int TimeInterval = 250;private StepCountListener mStepListeners;@Overridepublic void onSensorChanged(SensorEvent event) {for (int i = 0; i < 3; i++) {oriValues[i] = event.values[i];}gravityNew = (float) Math.sqrt(oriValues[0] * oriValues[0]+ oriValues[1] * oriValues[1] + oriValues[2] * oriValues[2]);detectorNewStep(gravityNew);}@Overridepublic void onAccuracyChanged(Sensor sensor, int accuracy) {//}public void initListener(StepCountListener listener) {this.mStepListeners = listener;}/** 检测步子，并开始计步* 1.传入sersor中的数据* 2.如果检测到了波峰，并且符合时间差以及阈值的条件，则判定为1步* 3.符合时间差条件，波峰波谷差值大于initialValue，则将该差值纳入阈值的计算中* */public void detectorNewStep(float values) {if (gravityOld == 0) {gravityOld = values;} else {if (detectorPeak(values, gravityOld)) {timeOfLastPeak = timeOfThisPeak;timeOfNow = System.currentTimeMillis();if (timeOfNow - timeOfLastPeak >= TimeInterval&& (peakOfWave - valleyOfWave >= ThreadValue)) {timeOfThisPeak = timeOfNow;/** 更新界面的处理，不涉及到算法* 一般在通知更新界面之前，增加下面处理，为了处理无效运动：* 1.连续记录10才开始计步* 2.例如记录的9步用户停住超过3秒，则前面的记录失效，下次从头开始* 3.连续记录了9步用户还在运动，之前的数据才有效* */mStepListeners.countStep();}if (timeOfNow - timeOfLastPeak >= TimeInterval&& (peakOfWave - valleyOfWave >= InitialValue)) {timeOfThisPeak = timeOfNow;ThreadValue = peakValleyThread(peakOfWave - valleyOfWave);}}}gravityOld = values;}/** 检测波峰* 以下四个条件判断为波峰：* 1.目前点为下降的趋势：isDirectionUp为false* 2.之前的点为上升的趋势：lastStatus为true* 3.到波峰为止，持续上升大于等于2次* 4.波峰值大于20* 记录波谷值* 1.观察波形图，可以发现在出现步子的地方，波谷的下一个就是波峰，有比较明显的特征以及差值* 2.所以要记录每次的波谷值，为了和下次的波峰做对比* */public boolean detectorPeak(float newValue, float oldValue) {lastStatus = isDirectionUp;if (newValue >= oldValue) {isDirectionUp = true;continueUpCount++;} else {continueUpFormerCount = continueUpCount;continueUpCount = 0;isDirectionUp = false;}if (!isDirectionUp && lastStatus&& (continueUpFormerCount >= 2 || oldValue >= 20)) {peakOfWave = oldValue;return true;} else if (!lastStatus && isDirectionUp) {valleyOfWave = oldValue;return false;} else {return false;}}/** 阈值的计算* 1.通过波峰波谷的差值计算阈值* 2.记录4个值，存入tempValue[]数组中* 3.在将数组传入函数averageValue中计算阈值* */public float peakValleyThread(float value) {float tempThread = ThreadValue;if (tempCount < ValueNum) {tempValue[tempCount] = value;tempCount++;} else {tempThread = averageValue(tempValue, ValueNum);for (int i = 1; i < ValueNum; i++) {tempValue[i - 1] = tempValue[i];}tempValue[ValueNum - 1] = value;}return tempThread;}/** 梯度化阈值* 1.计算数组的均值* 2.通过均值将阈值梯度化在一个范围里* */public float averageValue(float value[], int n) {float ave = 0;for (int i = 0; i < n; i++) {ave += value[i];}ave = ave / ValueNum;if (ave >= 8)ave = (float) 4.3;else if (ave >= 7 && ave < 8)ave = (float) 3.3;else if (ave >= 4 && ave < 7)ave = (float) 2.3;else if (ave >= 3 && ave < 4)ave = (float) 2.0;else {ave = (float) 1.3;}return ave;}}