我正在尝试使用AudioFileWriteBytes()将原始音频字节写入文件。这就是我正在做的事情:
void writeSingleChannelRingBufferDataToFileAsSInt16(AudioFileID audioFileID, AudioConverterRef audioConverter, ringBuffer *rb, SInt16 *holdingBuffer) {
// First, figure out which bits of audio we'll be
// writing to file from the ring buffer
UInt32 lastFreshSample = rb->lastWrittenIndex;
OSStatus status;
int numSamplesToWrite;
UInt32 numBytesToWrite;
if (lastFreshSample < rb->lastReadIndex) {
numSamplesToWrite = kNumPointsInWave + lastFreshSample - rb->lastReadIndex - 1;
}
else {
numSamplesToWrite = lastFreshSample - rb->lastReadIndex;
}
numBytesToWrite = numSamplesToWrite*sizeof(SInt16);
然后我们将音频数据(存储为浮点数)复制到将直接写入文件的保持缓冲区(SInt16)。复制看起来很时髦,因为它来自环形缓冲区。
UInt32 buffLen = rb->sizeOfBuffer - 1;
for (int i=0; i < numSamplesToWrite; ++i) {
holdingBuffer[i] = rb->data[(i + rb->lastReadIndex) & buffLen];
}
好的,现在我们实际上尝试将来自SInt16缓冲区“holdingBuffer”的音频写入音频文件。 NSLog将吐出错误-40,但也声称它正在写入字节。没有数据写入文件。
status = AudioFileWriteBytes(audioFileID, NO, 0, &numBytesToWrite, &holdingBuffer);
rb->lastReadIndex = lastFreshSample;
NSLog(@"Error = %d, wrote %d bytes", status, numBytesToWrite);
return;
这是什么错误-40?顺便说一句,如果我直接从ringBuffer写入文件,一切正常。当然这听起来像垃圾,因为我正在编写浮点数,而不是SInt16,但AudioFileWriteBytes并没有抱怨。
答案 0 :(得分:2)
关键是明确地将传入数据的字节顺序更改为big endian。我所要做的就是将CFSwapInt16HostToBig包装在我的数据周围以获得:
float audioVal = rb->data[(i + rb->lastReadIndex) & buffLen];
holdingBuffer[i] = CFSwapInt16HostToBig((SInt16) audioVal );