我在打电话：
mReadBytes = mAudioRecord.read(mBuffer, mConfig.frameSize());
在不需要额外处理的后台线程上。
AudioFormat.CHANNEL_IN_STEREO AudioFormat.ENCODING_PCM_16BIT 我正在尝试一次读取2048个字节[每个通道1024个字节/每个通道512个样本]。此数量应等于声音的 11609微秒：
11609 (msec) = 2048 (bytes) / (44100 (samples/sec) * 2 (bytes/sample/channel) * 2 (channels)) * 1000000 (msec/sec).
由于这个原因，我希望read()呼叫在这段时间过去之后最多返回（如果有的话，还要加上一些开销）。
当我使用MONO配置时，一切正常。我的假设是正确的-read()呼叫平均花费的时间（针对单个频道进行调整），通常少于该时间。
当我使用上面的STEREO配置时，read()调用花费的时间要比应有的要多 ，即使假设每次我调用记录器缓冲区为空时，延迟为< strong>巨大，并在最终文件中造成很大差距。
V/AudioMediaEncoder: read thread - reading took: 37.782295 should be: 11.609 delay: 26.173294
V/AudioMediaEncoder: read thread - reading took: 18.121565 should be: 11.609 delay: 6.5125647
V/AudioMediaEncoder: read thread - reading took: 18.233908 should be: 11.609 delay: 6.6249075
V/AudioMediaEncoder: read thread - reading took: 18.458492 should be: 11.609 delay: 6.849492
V/AudioMediaEncoder: read thread - reading took: 16.850992 should be: 11.609 delay: 5.241992
V/AudioMediaEncoder: read thread - reading took: 17.748648 should be: 11.609 delay: 6.1396475
V/AudioMediaEncoder: read thread - reading took: 19.446096 should be: 11.609 delay: 7.837096
V/AudioMediaEncoder: read thread - reading took: 35.627087 should be: 11.609 delay: 24.018085
V/AudioMediaEncoder: read thread - reading took: 16.654844 should be: 11.609 delay: 5.045844
V/AudioMediaEncoder: read thread - reading took: 19.425213 should be: 11.609 delay: 7.8162127
V/AudioMediaEncoder: read thread - reading took: 19.640837 should be: 11.609 delay: 8.0318365
V/AudioMediaEncoder: read thread - reading took: 15.437971 should be: 11.609 delay: 3.828971
V/AudioMediaEncoder: read thread - reading took: 18.271303 should be: 11.609 delay: 6.662303
V/AudioMediaEncoder: read thread - reading took: 19.074793 should be: 11.609 delay: 7.4657927
V/AudioMediaEncoder: read thread - reading took: 38.72766 should be: 11.609 delay: 27.11866
如您在上面看到的（以毫秒为单位的值），读取操作最多需要 40毫秒才能捕获声音缓冲区，该缓冲区的长度应 11.6毫秒。
不用说输出是垃圾。谁能解释原因或帮助？
AudioRecord内部缓冲区设置为2048 *25。更改此设置只会使所有情况变得更糟。
这是输出上述日志的代码：
long before = System.nanoTime();
mReadBytes = mAudioRecord.read(mCurrentBuffer, mConfig.frameSize());
long after = System.nanoTime();
float delayMillis = (after - before) / 1000000F;
float durationMillis = AudioTimestamp.bytesToMillis(mReadBytes, mConfig.byteRate());
Log.v(TAG, "read thread - reading took: " + delayMillis + " should be: " + durationMillis + " delay: " + (delayMillis - durationMillis));