使用 UWP 监控实时音频并检测枪声/拍手声
Posted
技术标签:
【中文标题】使用 UWP 监控实时音频并检测枪声/拍手声【英文标题】：Using UWP monitor live audio and detect gun-fire/clap sound 【发布时间】：2019-01-03 09:27:50 【问题描述】：我正在开发一个新的 UWP 应用程序，该应用程序应监控声音并为每次突然的敲击声触发一个事件（类似枪声或拍手声）。
它需要启用默认音频输入并监控实时音频。 设置音频灵敏度以识别环境噪音和识别拍手/枪声 当出现像拍手/枪声这样的高频声音时（理想情况下，它应该像配置的频率一样，比如 +/-40，那么它就是枪声/拍手）然后它应该调用一个事件。无需保存音频 我尝试实现this
SoundMonitoringPage：
public sealed partial class MyPage : Page
private async void Page_Loaded(object sender, RoutedEventArgs e)
string deviceId = Windows.Media.Devices.MediaDevice.GetDefaultAudioCaptureId(Windows.Media.Devices.AudioDeviceRole.Communications);
gameChatAudiostateMonitor = AudioStateMonitor.CreateForCaptureMonitoringWithCategoryAndDeviceId(MediaCategory.GameChat, deviceId);
gameChatAudioStateMonitor.SoundLevelChanged += GameChatSoundLevelChanged;
//other logic
声级变化：
private void GameChatSoundLevelChanged(AudioStateMonitor sender, object args)
switch (sender.SoundLevel)
case SoundLevel.Full:
LevelChangeEvent();
break;
case SoundLevel.Muted:
LevelChangeEvent();
break;
case SoundLevel.Low:
// Audio capture should never be "ducked", only muted or full volume.
Debug.WriteLine("Unexpected audio state change.");
break;
ENV：Windows 10 (v1809) IDE：VS 2017
不确定这是否是正确的方法。这不会启用音频，也不会触发级别更改事件。
我在 WinForms 和 NAudio 教程 here 中看到了其他选项。可能通过采样频率我可以检查事件...没有必须教程使用 NAudio 和 UWP 来绘制图形并识别频率。
更新：
遵循@Rob Caplan - MSFT 的建议，这就是我最终得到的结果
IMemoryBufferByteAccess.cs
// We are initializing a COM interface for use within the namespace
// This interface allows access to memory at the byte level which we need to populate audio data that is generated
[ComImport]
[Guid("5B0D3235-4DBA-4D44-865E-8F1D0E4FD04D")]
[InterfaceType(ComInterfaceType.InterfaceIsIUnknown)]
unsafe interface IMemoryBufferByteAccess
void GetBuffer(out byte* buffer, out uint capacity);
GunFireMonitorPage.xaml.cs
public sealed partial class GunFireMonitorPage : Page
private MainPage _rootPage;
public static GunFireMonitorPage Current;
private AudioGraph _graph;
private AudioDeviceOutputNode _deviceOutputNode;
private AudioFrameInputNode _frameInputNode;
public double Theta;
public DrivePage()
InitializeComponent();
Current = this;
protected override async void OnNavigatedTo(NavigationEventArgs e)
_rootPage = MainPage.Current;
await CreateAudioGraph();
protected override void OnNavigatedFrom(NavigationEventArgs e)
_graph?.Dispose();
private void Page_Loaded(object sender, RoutedEventArgs e)
private unsafe AudioFrame GenerateAudioData(uint samples)
// Buffer size is (number of samples) * (size of each sample)
// We choose to generate single channel (mono) audio. For multi-channel, multiply by number of channels
uint bufferSize = samples * sizeof(float);
AudioFrame audioFrame = new AudioFrame(bufferSize);
using (AudioBuffer buffer = audioFrame.LockBuffer(AudioBufferAccessMode.Write))
using (IMemoryBufferReference reference = buffer.CreateReference())
// Get the buffer from the AudioFrame
// ReSharper disable once SuspiciousTypeConversion.Global
// ReSharper disable once UnusedVariable
((IMemoryBufferByteAccess) reference).GetBuffer(out var dataInBytes, out var capacityInBytes);
// Cast to float since the data we are generating is float
var dataInFloat = (float*)dataInBytes;
float freq = 1000; // choosing to generate frequency of 1kHz
float amplitude = 0.3f;
int sampleRate = (int)_graph.EncodingProperties.SampleRate;
double sampleIncrement = (freq * (Math.PI * 2)) / sampleRate;
// Generate a 1kHz sine wave and populate the values in the memory buffer
for (int i = 0; i < samples; i++)
double sinValue = amplitude * Math.Sin(Theta);
dataInFloat[i] = (float)sinValue;
Theta += sampleIncrement;
return audioFrame;
private void node_QuantumStarted(AudioFrameInputNode sender, FrameInputNodeQuantumStartedEventArgs args)
// GenerateAudioData can provide PCM audio data by directly synthesizing it or reading from a file.
// Need to know how many samples are required. In this case, the node is running at the same rate as the rest of the graph
// For minimum latency, only provide the required amount of samples. Extra samples will introduce additional latency.
uint numSamplesNeeded = (uint)args.RequiredSamples;
if (numSamplesNeeded != 0)
AudioFrame audioData = GenerateAudioData(numSamplesNeeded);
_frameInputNode.AddFrame(audioData);
private void Button_Click(object sender, RoutedEventArgs e)
if (generateButton.Content != null && generateButton.Content.Equals("Generate Audio"))
_frameInputNode.Start();
generateButton.Content = "Stop";
audioPipe.Fill = new SolidColorBrush(Colors.Blue);
else if (generateButton.Content != null && generateButton.Content.Equals("Stop"))
_frameInputNode.Stop();
generateButton.Content = "Generate Audio";
audioPipe.Fill = new SolidColorBrush(Color.FromArgb(255, 49, 49, 49));
private async Task CreateAudioGraph()
// Create an AudioGraph with default settings
AudioGraphSettings settings = new AudioGraphSettings(AudioRenderCategory.Media);
CreateAudioGraphResult result = await AudioGraph.CreateAsync(settings);
if (result.Status != AudioGraphCreationStatus.Success)
// Cannot create graph
_rootPage.NotifyUser($"AudioGraph Creation Error because result.Status.ToString()", NotifyType.ErrorMessage);
return;
_graph = result.Graph;
// Create a device output node
CreateAudioDeviceOutputNodeResult deviceOutputNodeResult = await _graph.CreateDeviceOutputNodeAsync();
if (deviceOutputNodeResult.Status != AudioDeviceNodeCreationStatus.Success)
// Cannot create device output node
_rootPage.NotifyUser(
$"Audio Device Output unavailable because deviceOutputNodeResult.Status.ToString()", NotifyType.ErrorMessage);
speakerContainer.Background = new SolidColorBrush(Colors.Red);
_deviceOutputNode = deviceOutputNodeResult.DeviceOutputNode;
_rootPage.NotifyUser("Device Output Node successfully created", NotifyType.StatusMessage);
speakerContainer.Background = new SolidColorBrush(Colors.Green);
// Create the FrameInputNode at the same format as the graph, except explicitly set mono.
AudioEncodingProperties nodeEncodingProperties = _graph.EncodingProperties;
nodeEncodingProperties.ChannelCount = 1;
_frameInputNode = _graph.CreateFrameInputNode(nodeEncodingProperties);
_frameInputNode.AddOutgoingConnection(_deviceOutputNode);
frameContainer.Background = new SolidColorBrush(Colors.Green);
// Initialize the Frame Input Node in the stopped state
_frameInputNode.Stop();
// Hook up an event handler so we can start generating samples when needed
// This event is triggered when the node is required to provide data
_frameInputNode.QuantumStarted += node_QuantumStarted;
// Start the graph since we will only start/stop the frame input node
_graph.Start();
GunFireMonitorPage.xaml
<Page
x:Class="SmartPileInspector.xLite.GunFireMonitorPage"
xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
xmlns:d="http://schemas.microsoft.com/expression/blend/2008"
xmlns:mc="http://schemas.openxmlformats.org/markup-compatibility/2006"
mc:Ignorable="d" Loaded="Page_Loaded"
HorizontalAlignment="Center"
Background="ThemeResource ApplicationPageBackgroundThemeBrush">
<ScrollViewer HorizontalAlignment="Center">
<StackPanel HorizontalAlignment="Center">
<!-- more page content -->
<Grid HorizontalAlignment="Center">
<Grid.ColumnDefinitions>
<ColumnDefinition Width="*"/>
<ColumnDefinition Width="*"/>
</Grid.ColumnDefinitions>
<Grid.RowDefinitions>
<RowDefinition Height="55"></RowDefinition>
</Grid.RowDefinitions>
</Grid>
<AppBarButton x:Name="generateButton" Content="Generate Audio" Click="Button_Click" MinWidth="120" MinHeight="45" Margin="0,50,0,0"/>
<Border x:Name="frameContainer" BorderThickness="0" Background="#4A4A4A" MinWidth="120" MinHeight="45" Margin="0,20,0,0">
<TextBlock x:Name="frame" Text="Frame Input" VerticalAlignment="Center" HorizontalAlignment="Center" />
</Border>
<StackPanel>
<Rectangle x:Name="audioPipe" Margin="0,20,0,0" Height="10" MinWidth="160" Fill="#313131" HorizontalAlignment="Stretch"/>
</StackPanel>
<Border x:Name="speakerContainer" BorderThickness="0" Background="#4A4A4A" MinWidth="120" MinHeight="45" Margin="0,20,0,0">
<TextBlock x:Name="speaker" Text="Output Device" VerticalAlignment="Center" HorizontalAlignment="Center" />
</Border>
<!--</AppBar>-->
</StackPanel>
</ScrollViewer>
</Page>
没有生成图表。并且有蓝线连续的哔声。 非常感谢任何帮助
更新：实现 AudioVisualizer
在AudioVisualizer 的帮助下，我能够绘制虱子音频图。
AudioGraph _graph;
AudioDeviceInputNode _inputNode;
PlaybackSource _source;
SourceConverter _converter;
protected override void OnNavigatedTo(NavigationEventArgs e)
_rootPage = MainPage.Current;
_rootPage.SetDimensions(700, 600);
base.OnNavigatedTo(e);
CreateAudioGraphAsync();
protected override void OnNavigatedFrom(NavigationEventArgs e)
base.OnNavigatedFrom(e);
_graph?.Stop();
_graph?.Dispose();
_graph = null;
async void CreateAudioGraphAsync()
var graphResult = await AudioGraph.CreateAsync(new AudioGraphSettings(Windows.Media.Render.AudioRenderCategory.Media));
if (graphResult.Status != AudioGraphCreationStatus.Success)
throw new InvalidOperationException($"Graph creation failed graphResult.Status");
_graph = graphResult.Graph;
var inputNodeResult = await _graph.CreateDeviceInputNodeAsync(MediaCategory.Media);
if (inputNodeResult.Status == AudioDeviceNodeCreationStatus.Success)
_inputNode = inputNodeResult.DeviceInputNode;
_source = PlaybackSource.CreateFromAudioNode(_inputNode);
_converter = new SourceConverter
Source = _source.Source,
MinFrequency = 110.0f,
MaxFrequency = 3520.0f,
FrequencyCount = 12 * 5 * 5,
FrequencyScale = ScaleType.Linear,
SpectrumRiseTime = TimeSpan.FromMilliseconds(20),
SpectrumFallTime = TimeSpan.FromMilliseconds(200),
RmsRiseTime = TimeSpan.FromMilliseconds(20),
RmsFallTime = TimeSpan.FromMilliseconds(500),
ChannelCount = 1
;
// Note A2
// Note A7
// 5 octaves, 5 bars per note
// Use RMS to gate noise, fast rise slow fall
NotesSpectrum.Source = _converter;
_graph.Start();
else
_rootPage.NotifyUser("Cannot access microphone", NotifyType.ErrorMessage);
现在的挑战是当波频率高于阈值时如何连接事件？在那种情况下，我想计算拍摄次数、时间戳和强度。
示例声音
这是我的Recording of live sound，你可以在这里，当大锤敲击（每秒或更少）时，我想调用一个事件。
【问题讨论】：
【参考方案1】：您可以通过查找该帧中所有 pcm 数据的平均振幅来找到该帧的分贝。我相信您想要创建一个处理输入的图表，看起来像这样
private static event LoudNoise<double>;
private static int quantum = 0;
static AudioGraph ingraph;
private static AudioDeviceInputNode deviceInputNode;
private static AudioFrameOutputNode frameOutputNode;
public static async Task<bool> CreateInputDeviceNode(string deviceId)
Console.WriteLine("Creating AudioGraphs");
// Create an AudioGraph with default settings
AudioGraphSettings graphsettings = new AudioGraphSettings(AudioRenderCategory.Media);
graphsettings.EncodingProperties = new AudioEncodingProperties();
graphsettings.EncodingProperties.Subtype = "Float";
graphsettings.EncodingProperties.SampleRate = 48000;
graphsettings.EncodingProperties.ChannelCount = 2;
graphsettings.EncodingProperties.BitsPerSample = 32;
graphsettings.EncodingProperties.Bitrate = 3072000;
//settings.DesiredSamplesPerQuantum = 960;
//settings.QuantumSizeSelectionMode = QuantumSizeSelectionMode.ClosestToDesired;
CreateAudioGraphResult graphresult = await AudioGraph.CreateAsync(graphsettings);
if (graphresult.Status != AudioGraphCreationStatus.Success)
// Cannot create graph
return false;
ingraph = graphresult.Graph;AudioGraphSettings nodesettings = new AudioGraphSettings(AudioRenderCategory.GameChat);
nodesettings.EncodingProperties = AudioEncodingProperties.CreatePcm(48000, 2, 32);
nodesettings.DesiredSamplesPerQuantum = 960;
nodesettings.QuantumSizeSelectionMode = QuantumSizeSelectionMode.ClosestToDesired;
frameOutputNode = ingraph.CreateFrameOutputNode(ingraph.EncodingProperties);
quantum = 0;
ingraph.QuantumStarted += Graph_QuantumStarted;
DeviceInformation selectedDevice;
string device = Windows.Media.Devices.MediaDevice.GetDefaultAudioCaptureId(Windows.Media.Devices.AudioDeviceRole.Default);
if (!string.IsNullOrEmpty(device))
selectedDevice = await DeviceInformation.CreateFromIdAsync(device);
else
return false;
CreateAudioDeviceInputNodeResult result =
await ingraph.CreateDeviceInputNodeAsync(MediaCategory.Media, nodesettings.EncodingProperties, selectedDevice);
if (result.Status != AudioDeviceNodeCreationStatus.Success)
// Cannot create device output node
return false;
deviceInputNode = result.DeviceInputNode;
deviceInputNode.AddOutgoingConnection(frameOutputNode);
frameOutputNode.Start();
ingraph.Start();
return true;
private static void Graph_QuantumStarted(AudioGraph sender, object args)
if (++quantum % 2 == 0)
AudioFrame frame = frameOutputNode.GetFrame();
float[] dataInFloats;
using (AudioBuffer buffer = frame.LockBuffer(AudioBufferAccessMode.Write))
using (IMemoryBufferReference reference = buffer.CreateReference())
// Get the buffer from the AudioFrame
((IMemoryBufferByteAccess)reference).GetBuffer(out byte* dataInBytes, out uint capacityInBytes);
float* dataInFloat = (float*)dataInBytes;
dataInFloats = new float[capacityInBytes / sizeof(float)];
for (int i = 0; i < capacityInBytes / sizeof(float); i++)
dataInFloats[i] = dataInFloat[i];
double decibels = 0f;
foreach (var sample in dataInFloats)
decibels += Math.Abs(sample);
decibels = 20 * Math.Log10(decibels / dataInFloats.Length);
// You can pass the decibel value where ever you'd like from here
if (decibels > 10)
LoudNoise?.Invoke(this, decibels);
附：我做了所有这些静态的，但如果它们都在同一个实例中自然会起作用
我还从我自己的项目中部分复制了这个，所以它可能有一些我忘记修剪的部分。希望对你有帮助
【讨论】：
让我们continue this discussion in chat。 在我的 Xaml 页面 imgur.com/a/tJZXowO 上的 TextBox 控件上设置值会引发错误@ 它试图从不是 UI 线程的线程中设置值。您需要在await Windows.UI.Xaml.Dispatcher.RunAsync(Windows.UI.Core.CoreDispatcherPriority.Normal, async () => /*Your Code*/); 中运行您的代码【参考方案2】：
回答“这是正确的方法”问题：不，AudioStateMonitor 对解决问题没有帮助。
AudioStateMonitor.SoundLevelChanged 会告诉您系统是否正在闪避您的声音，以免干扰其他东西。例如，它可以使音乐静音以支持电话铃声。 SoundLevelChanged 不会告诉您任何有关录制声音的音量或频率的信息，而这正是您检测拍手所需要的。
正确的方法是使用 AudioGraph（或 WASAPI，但不是来自 C#）将原始音频捕获到 AudioFrameOutputNode 以处理信号，然后通过 FFT 运行以检测您的声音目标频率和数量。 AudioCreation sample 演示了使用 AudioGraph，但不具体是 AudioFrameOutputNode。
每https://home.howstuffworks.com/clapper1.htm 拍手将在 2200Hz 到 2800Hz 的频率范围内。
识别枪声看起来要复杂得多，不同的枪具有非常不同的特征。快速搜索发现了几篇关于这个而不是琐碎算法的研究论文。我怀疑你会想要某种机器学习来对这些进行分类。这是之前讨论使用 ML 区分枪声和非枪声的帖子：SVM for one Vs all acoustic signal classification
【讨论】：
以上是关于使用 UWP 监控实时音频并检测枪声/拍手声的主要内容，如果未能解决你的问题，请参考以下文章
如何检测其他应用程序的音频是不是在后台播放？ （UWP，Windows 10）