AMR 安卓编码 https://github.com/kevinho/opencore-amr-Android
编码float[] to short[] 互转 http://stackoverflow.com/questions/16529857/converting-16-bit-short-to-32-bit-float
因为编码采样的到的浮点数是 -1.0f ~ 1.0f 而 signed short is -32767 <= Xn <= +32767, 所以将浮点数转化short 乘32767
这里编码转换如果是将float拆分为2个short 那么的到的采样值是不对的
https://github.com/fholm/unityassets/tree/master/VoiceChat
https://github.com/Wave-Unity-Library/Motion-Mapping