Android使用AudioRecord判斷是否有音頻輸入

2019-12-12 04:02:39

字體：大中小

來源：轉載

供稿：網友

Android黨都應該玩過一個叫吹裙子的游戲，這個游戲就是原理就是通過監聽用戶吹出的氣的力度來決定如何把MM的裙子弄飛起來的，所以關鍵在于如何判斷用戶吹氣的力度問題。現在公司剛好有這個需求要評估就是需要一直監聽用戶的語音輸入，當在兩秒內沒有語音輸入時候就暫停某項事情，有語音輸入的時候就要繼續做某件事。其實這兩件事情的本質原理是一樣的，就是通過這AudioRecord來處理用戶輸入的原始音頻數據，從而計算出當前用戶輸入的音量大小來判斷用戶是否有語音輸入。下面貼上一段代碼用戶可以自行研究。

public static int calculateVolume(byte[] var0, int var1) {    int[] var3 = null;    int var4 = var0.length;    int var2;    if(var1 == 8) {      var3 = new int[var4];      for(var2 = 0; var2 < var4; ++var2) {        var3[var2] = var0[var2];      }    } else if(var1 == 16) {      var3 = new int[var4 / 2];      for(var2 = 0; var2 < var4 / 2; ++var2) {        byte var5 = var0[var2 * 2];        byte var6 = var0[var2 * 2 + 1];        int var13;        if(var5 < 0) {          var13 = var5 + 256;        } else {          var13 = var5;        }        short var7 = (short)(var13 + 0);        if(var6 < 0) {          var13 = var6 + 256;        } else {          var13 = var6;        }        var3[var2] = (short)(var7 + (var13 << 8));      }    }    int[] var8 = var3;    if(var3 != null && var3.length != 0) {      float var10 = 0.0F;      for(int var11 = 0; var11 < var8.length; ++var11) {        var10 += (float)(var8[var11] * var8[var11]);      }      var10 /= (float)var8.length;      float var12 = 0.0F;      for(var4 = 0; var4 < var8.length; ++var4) {        var12 += (float)var8[var4];      }      var12 /= (float)var8.length;      var4 = (int)(Math.pow(2.0D, (double)(var1 - 1)) - 1.0D);      double var14 = Math.sqrt((double)(var10 - var12 * var12));      int var9;      if((var9 = (int)(10.0D * Math.log10(var14 * 10.0D * Math.sqrt(2.0D) / (double)var4 + 1.0D))) < 0) {        var9 = 0;      }      if(var9 > 10) {        var9 = 10;      }      return var9;    } else {      return 0;    }}

上述方法的形參中的一個字節數組便是當前捕捉到用戶輸入的原始音頻數據，第二個形參則是音頻數據格式位PCM，16位每個樣本，比如如果音頻數據格式設為AudioFormat.ENCODING_PCM_16BIT，則參數傳16。最后實現的效果就是如下：

01-12 18:54:17.222 E/recordtest(24249): no voice input
…………….
01-12 18:54:19.027 E/recordtest(24249): no voice input
01-12 18:54:19.067 E/recordtest(24249): no voice input
01-12 18:54:19.105 E/recordtest(24249): no voice input
01-12 18:54:19.145 E/recordtest(24249): no voice input
01-12 18:54:19.188 E/recordtest(24249): no voice input
01-12 18:54:19.227 E/recordtest(24249): no voice input
01-12 18:54:19.228 E/recordtest(24249): no voice input in 2 sec
01-12 18:54:19.269 E/recordtest(24249): no voice input
01-12 18:54:19.306 E/recordtest(24249): no voice input
01-12 18:54:19.345 E/recordtest(24249): no voice input
01-12 18:54:19.387 E/recordtest(24249): no voice input
01-12 18:54:19.427 E/recordtest(24249): no voice input
01-12 18:54:19.466 E/recordtest(24249): has voice input
01-12 18:54:19.505 E/recordtest(24249): has voice input
01-12 18:54:19.546 E/recordtest(24249): has voice input
01-12 18:54:19.586 E/recordtest(24249): has voice input
01-12 18:54:19.646 E/recordtest(24249): has voice input
01-12 18:54:19.686 E/recordtest(24249): has voice input
01-12 18:54:19.726 E/recordtest(24249): has voice input
01-12 18:54:19.766 E/recordtest(24249): has voice input