正如标题所示，我希望从 .NET Core 中的麦克风捕获音频。对我来说重要的是它是跨平台的。我已经在 Windows、Linux 和 OSX 上的 OpenCV 中进行了大量视频处理。音频对我来说是一个缺失的部分。
【问题讨论】：
标签： .net audio .net-core cross-platform
OpenTK.NETCore 是一个非常好的 OpenTK 跨平台端口。
使用它的 AudioCapture 类，我已经能够跨不同的操作系统平台捕获麦克风数据。
using (var audioCapture = new AudioCapture(_recorders.First(), _sampling_rate, ALFormat.Mono16, buffer_length_samples))
{
audioCapture.Start();
int available_samples = audioCapture.AvailableSamples;
_buffer = _pool.Rent(MathHelper.NextPowerOfTwo((int)((available_samples * _sampleToByte / (double)BlittableValueType.StrideOf(_buffer)) + 0.5)));
if (available_samples > 0)
{
audioCapture.ReadSamples(_buffer, available_samples);
}
else
{
_pool.Return(_buffer);
}
}
编辑：基于 OpenTK 示例中的 Parrot 项目。
【讨论】：
很遗憾，我不知道有什么 API 或库可以为您做到这一点。
如果其他人也不知道，那么你有两种选择：
【讨论】：
我之前在 .NET Core 上编写过一个记录器应用程序，但它不再是我的财产了……但是，这是我在处理它时使用的 the library。
LibsoundIO Sharp 能够将配置扩展到诸如输入或输出的采样率之类的东西，并且有这样的好例子：
class Record
{
static SoundIORingBuffer ring_buffer = null;
static SoundIOFormat [] prioritized_formats = {
SoundIODevice.Float32NE,
SoundIODevice.Float32FE,
SoundIODevice.S32NE,
SoundIODevice.S32FE,
SoundIODevice.S24NE,
SoundIODevice.S24FE,
SoundIODevice.S16NE,
SoundIODevice.S16FE,
SoundIODevice.Float64NE,
SoundIODevice.Float64FE,
SoundIODevice.U32NE,
SoundIODevice.U32FE,
SoundIODevice.U24NE,
SoundIODevice.U24FE,
SoundIODevice.U16NE,
SoundIODevice.U16FE,
SoundIOFormat.S8,
SoundIOFormat.U8,
SoundIOFormat.Invalid,
};
static readonly int [] prioritized_sample_rates = {
48000,
44100,
96000,
24000,
0,
};
public static int Main (string [] args)
{
string device_id = null;
string backend_name = null;
bool raw = false;
string outfile = null;
foreach (var arg in args) {
switch (arg) {
case "--raw":
raw = true;
continue;
default:
if (arg.StartsWith ("--backend:"))
backend_name = arg.Substring (arg.IndexOf (':') + 1);
else if (arg.StartsWith ("--device:"))
device_id = arg.Substring (arg.IndexOf (':') + 1);
else
outfile = arg;
continue;
}
}
var api = new SoundIO ();
var backend = backend_name == null ? SoundIOBackend.None : (SoundIOBackend)Enum.Parse (typeof (SoundIOBackend), backend_name);
if (backend == SoundIOBackend.None)
api.Connect ();
else
api.ConnectBackend (backend);
Console.WriteLine ("backend: " + api.CurrentBackend);
api.FlushEvents ();
var device = device_id == null ? api.GetInputDevice (api.DefaultInputDeviceIndex) :
Enumerable.Range (0, api.InputDeviceCount)
.Select (i => api.GetInputDevice (i))
.FirstOrDefault (d => d.Id == device_id && d.IsRaw == raw);
if (device == null) {
Console.Error.WriteLine ("device " + device_id + " not found.");
return 1;
}
Console.WriteLine ("device: " + device.Name);
if (device.ProbeError != 0) {
Console.Error.WriteLine ("Cannot probe device " + device_id + ".");
return 1;
}
var sample_rate = prioritized_sample_rates.First (sr => device.SupportsSampleRate (sr));
var fmt = prioritized_formats.First (f => device.SupportsFormat (f));
var instream = device.CreateInStream ();
instream.Format = fmt;
instream.SampleRate = sample_rate;
instream.ReadCallback = (fmin, fmax) => read_callback (instream, fmin, fmax);
instream.OverflowCallback = () => overflow_callback (instream);
instream.Open ();
const int ring_buffer_duration_seconds = 30;
int capacity = (int)(ring_buffer_duration_seconds * instream.SampleRate * instream.BytesPerFrame);
ring_buffer = api.CreateRingBuffer (capacity);
var buf = ring_buffer.WritePointer;
instream.Start ();
Console.WriteLine ("Type CTRL+C to quit by killing process...");
using (var fs = File.OpenWrite (outfile)) {
var arr = new byte [capacity];
unsafe {
fixed (void* arrptr = arr) {
for (; ; ) {
api.FlushEvents ();
Thread.Sleep (1000);
int fill_bytes = ring_buffer.FillCount;
var read_buf = ring_buffer.ReadPointer;
Buffer.MemoryCopy ((void*)read_buf, arrptr, fill_bytes, fill_bytes);
fs.Write (arr, 0, fill_bytes);
ring_buffer.AdvanceReadPointer (fill_bytes);
}
}
}
}
instream.Dispose ();
device.RemoveReference ();
api.Dispose ();
return 0;
}
static void read_callback (SoundIOInStream instream, int frame_count_min, int frame_count_max)
{
var write_ptr = ring_buffer.WritePointer;
int free_bytes = ring_buffer.FreeCount;
int free_count = free_bytes / instream.BytesPerFrame;
if (frame_count_min > free_count)
throw new InvalidOperationException ("ring buffer overflow"); // panic()
int write_frames = Math.Min (free_count, frame_count_max);
int frames_left = write_frames;
for (; ; ) {
int frame_count = frames_left;
var areas = instream.BeginRead (ref frame_count);
if (frame_count == 0)
break;
if (areas.IsEmpty) {
// Due to an overflow there is a hole. Fill the ring buffer with
// silence for the size of the hole.
for (int i = 0; i < frame_count * instream.BytesPerFrame; i++)
Marshal.WriteByte (write_ptr + i, 0);
Console.Error.WriteLine ("Dropped {0} frames due to internal overflow", frame_count);
} else {
for (int frame = 0; frame < frame_count; frame += 1) {
int chCount = instream.Layout.ChannelCount;
int copySize = instream.BytesPerSample;
unsafe {
for (int ch = 0; ch < chCount; ch += 1) {
var area = areas.GetArea (ch);
Buffer.MemoryCopy ((void*)area.Pointer, (void*)write_ptr, copySize, copySize);
area.Pointer += area.Step;
write_ptr += copySize;
}
}
}
}
instream.EndRead ();
frames_left -= frame_count;
if (frames_left <= 0)
break;
}
int advance_bytes = write_frames * instream.BytesPerFrame;
ring_buffer.AdvanceWritePointer (advance_bytes);
}
static int overflow_callback_count = 0;
static void overflow_callback (SoundIOInStream instream)
{
Console.Error.WriteLine ("overflow {0}", overflow_callback_count++);
}
}
Adapted from sio-record/Program.cs
就 .NET Core 而言，您可以将其作为独立项目留在单例类中，或者如果您正在处理具有 DI 容器的项目，只需将其作为单例放入。
话虽如此，您需要libsoundio 才能使事情正常进行，因为毕竟它就像一个用于音频的“java”，而 libsoundio-sharp 是 C# 的包装器。干杯！
【讨论】：