如何使用 JS Web Audio API 进行节拍检测？

Question

我有兴趣使用 JavaScript WebAudioAPI 检测歌曲节拍，然后在画布中呈现它们。

我可以处理画布部分，但我不是音频大佬，真的不明白如何用 JavaScript 制作节拍检测器。

我已经尝试关注this article，但在我的一生中，无法将每个函数之间的点连接起来制作一个函数式程序。

我知道我应该给你看一些代码，但老实说我没有，我所有的尝试都失败了，相关代码在前面提到的文章中。

无论如何，我真的很感激一些指导，或者更好的演示如何使用 WebAudioAPI 实际检测歌曲节拍。

谢谢！

Answer 1

要了解the referenced article by Joe Sullivan 的主要内容是，尽管它提供了很多源代码，但它远非最终和完整的代码。要获得可行的解决方案，您仍然需要一些编码和调试技能。

这个答案的大部分代码来自引用的文章，原始许可适用于适当的地方。

下面是使用上述文章描述的函数的简单示例实现，您仍然需要找出函数解决方案的正确阈值。

---

代码由为答案编写的准备代码组成：

• 通过the FileReader API读取本地文件
• 使用the AudioContext API将文件解码为音频数据

然后，如文章中所述：

• 过滤音频，在本例中使用low-pass filter
• 使用阈值计算峰值
• 分组间隔计数，然后是速度计数

对于阈值，我使用了介于最大值和最小值之间的任意值 0.98；在分组时，我添加了一些额外的检查和任意舍入以避免可能的无限循环并使其成为易于调试的示例。

请注意，为了保持示例实现的简洁，评论很少，因为：

• 处理背后的逻辑在参考文章中进行了解释
• 语法可以参考相关方法的API文档

---

audio_file.onchange = function() {
  var file = this.files[0];
  var reader = new FileReader();
  var context = new(window.AudioContext || window.webkitAudioContext)();
  reader.onload = function() {
    context.decodeAudioData(reader.result, function(buffer) {
      prepare(buffer);
    });
  };
  reader.readAsArrayBuffer(file);
};

function prepare(buffer) {
  var offlineContext = new OfflineAudioContext(1, buffer.length, buffer.sampleRate);
  var source = offlineContext.createBufferSource();
  source.buffer = buffer;
  var filter = offlineContext.createBiquadFilter();
  filter.type = "lowpass";
  source.connect(filter);
  filter.connect(offlineContext.destination);
  source.start(0);
  offlineContext.startRendering();
  offlineContext.oncomplete = function(e) {
    process(e);
  };
}

function process(e) {
  var filteredBuffer = e.renderedBuffer;
  //If you want to analyze both channels, use the other channel later
  var data = filteredBuffer.getChannelData(0);
  var max = arrayMax(data);
  var min = arrayMin(data);
  var threshold = min + (max - min) * 0.98;
  var peaks = getPeaksAtThreshold(data, threshold);
  var intervalCounts = countIntervalsBetweenNearbyPeaks(peaks);
  var tempoCounts = groupNeighborsByTempo(intervalCounts);
  tempoCounts.sort(function(a, b) {
    return b.count - a.count;
  });
  if (tempoCounts.length) {
    output.innerHTML = tempoCounts[0].tempo;
  }
}

// http://tech.beatport.com/2014/web-audio/beat-detection-using-web-audio/
function getPeaksAtThreshold(data, threshold) {
  var peaksArray = [];
  var length = data.length;
  for (var i = 0; i < length;) {
    if (data[i] > threshold) {
      peaksArray.push(i);
      // Skip forward ~ 1/4s to get past this peak.
      i += 10000;
    }
    i++;
  }
  return peaksArray;
}

function countIntervalsBetweenNearbyPeaks(peaks) {
  var intervalCounts = [];
  peaks.forEach(function(peak, index) {
    for (var i = 0; i < 10; i++) {
      var interval = peaks[index + i] - peak;
      var foundInterval = intervalCounts.some(function(intervalCount) {
        if (intervalCount.interval === interval) return intervalCount.count++;
      });
      //Additional checks to avoid infinite loops in later processing
      if (!isNaN(interval) && interval !== 0 && !foundInterval) {
        intervalCounts.push({
          interval: interval,
          count: 1
        });
      }
    }
  });
  return intervalCounts;
}

function groupNeighborsByTempo(intervalCounts) {
  var tempoCounts = [];
  intervalCounts.forEach(function(intervalCount) {
    //Convert an interval to tempo
    var theoreticalTempo = 60 / (intervalCount.interval / 44100);
    theoreticalTempo = Math.round(theoreticalTempo);
    if (theoreticalTempo === 0) {
      return;
    }
    // Adjust the tempo to fit within the 90-180 BPM range
    while (theoreticalTempo < 90) theoreticalTempo *= 2;
    while (theoreticalTempo > 180) theoreticalTempo /= 2;

    var foundTempo = tempoCounts.some(function(tempoCount) {
      if (tempoCount.tempo === theoreticalTempo) return tempoCount.count += intervalCount.count;
    });
    if (!foundTempo) {
      tempoCounts.push({
        tempo: theoreticalTempo,
        count: intervalCount.count
      });
    }
  });
  return tempoCounts;
}

// http://stackoverflow.com/questions/1669190/javascript-min-max-array-values
function arrayMin(arr) {
  var len = arr.length,
    min = Infinity;
  while (len--) {
    if (arr[len] < min) {
      min = arr[len];
    }
  }
  return min;
}

function arrayMax(arr) {
  var len = arr.length,
    max = -Infinity;
  while (len--) {
    if (arr[len] > max) {
      max = arr[len];
    }
  }
  return max;
}

<input id="audio_file" type="file" accept="audio/*"></input>
<audio id="audio_player"></audio>
<p>
  Most likely tempo: <span id="output"></span>
</p>

Answer 2

我在这里写了一个教程，展示了如何使用 javascript Web Audio API 做到这一点。

https://askmacgyver.com/blog/tutorial/how-to-implement-tempo-detection-in-your-application

步骤大纲

1. 将音频文件转换为数组缓冲区
2. 通过低通滤波器运行阵列缓冲区
3. 从数组缓冲区修剪 10 秒剪辑
4. 下采样数据
5. 规范化数据
6. 统计卷分组
7. 从分组计数推断节奏

下面的这段代码完成了繁重的工作。

将音频文件加载到数组缓冲区并通过低通滤波器

function createBuffers(url) {

 // Fetch Audio Track via AJAX with URL
 request = new XMLHttpRequest();

 request.open('GET', url, true);
 request.responseType = 'arraybuffer';

 request.onload = function(ajaxResponseBuffer) {

    // Create and Save Original Buffer Audio Context in 'originalBuffer'
    var audioCtx = new AudioContext();
    var songLength = ajaxResponseBuffer.total;

    // Arguments: Channels, Length, Sample Rate
    var offlineCtx = new OfflineAudioContext(1, songLength, 44100);
    source = offlineCtx.createBufferSource();
    var audioData = request.response;
    audioCtx.decodeAudioData(audioData, function(buffer) {

         window.originalBuffer = buffer.getChannelData(0);
         var source = offlineCtx.createBufferSource();
         source.buffer = buffer;

         // Create a Low Pass Filter to Isolate Low End Beat
         var filter = offlineCtx.createBiquadFilter();
         filter.type = "lowpass";
         filter.frequency.value = 140;
         source.connect(filter);
         filter.connect(offlineCtx.destination);

            // Render this low pass filter data to new Audio Context and Save in 'lowPassBuffer'
            offlineCtx.startRendering().then(function(lowPassAudioBuffer) {

             var audioCtx = new(window.AudioContext || window.webkitAudioContext)();
             var song = audioCtx.createBufferSource();
             song.buffer = lowPassAudioBuffer;
             song.connect(audioCtx.destination);

             // Save lowPassBuffer in Global Array
             window.lowPassBuffer = song.buffer.getChannelData(0);
             console.log("Low Pass Buffer Rendered!");
            });

        },
        function(e) {});
 }
 request.send();
}


createBuffers('https://askmacgyver.com/test/Maroon5-Moves-Like-Jagger-128bpm.mp3');

您现在有一个低通滤波歌曲（和原始）的数组缓冲区

它由多个条目组成，sampleRate（44100 乘以歌曲的秒数）。

window.lowPassBuffer  // Low Pass Array Buffer
window.originalBuffer // Original Non Filtered Array Buffer

从歌曲中剪掉一个 10 秒的片段

function getClip(length, startTime, data) {

  var clip_length = length * 44100;
  var section = startTime * 44100;
  var newArr = [];

  for (var i = 0; i < clip_length; i++) {
     newArr.push(data[section + i]);
  }

  return newArr;
}

// Overwrite our array buffer to a 10 second clip starting from 00:10s
window.lowPassFilter = getClip(10, 10, lowPassFilter);

对您的剪辑进行下采样

function getSampleClip(data, samples) {

  var newArray = [];
  var modulus_coefficient = Math.round(data.length / samples);

  for (var i = 0; i < data.length; i++) {
     if (i % modulus_coefficient == 0) {
         newArray.push(data[i]);
     }
  }
  return newArray;
}

// Overwrite our array to down-sampled array.
lowPassBuffer = getSampleClip(lowPassFilter, 300);

规范化您的数据

function normalizeArray(data) {

 var newArray = [];

 for (var i = 0; i < data.length; i++) {
     newArray.push(Math.abs(Math.round((data[i + 1] - data[i]) * 1000)));
 }

 return newArray;
}

// Overwrite our array to the normalized array
lowPassBuffer = normalizeArray(lowPassBuffer);

计算扁线分组

function countFlatLineGroupings(data) {

 var groupings = 0;
 var newArray = normalizeArray(data);

 function getMax(a) {
    var m = -Infinity,
        i = 0,
        n = a.length;

    for (; i != n; ++i) {
        if (a[i] > m) {
            m = a[i];
        }
    }
    return m;
 }

 function getMin(a) {
    var m = Infinity,
        i = 0,
        n = a.length;

    for (; i != n; ++i) {
        if (a[i] < m) {
            m = a[i];
        }
    }
    return m;
 }

 var max = getMax(newArray);
 var min = getMin(newArray);
 var count = 0;
 var threshold = Math.round((max - min) * 0.2);

 for (var i = 0; i < newArray.length; i++) {

   if (newArray[i] > threshold && newArray[i + 1] < threshold && newArray[i + 2] < threshold && newArray[i + 3] < threshold && newArray[i + 6] < threshold) {
        count++;
    }
 }

 return count;
}

// Count the Groupings
countFlatLineGroupings(lowPassBuffer);

将 10 秒分组计数缩放为 60 秒以得出每分钟节拍数

var final_tempo = countFlatLineGroupings(lowPassBuffer);

// final_tempo will be 21
final_tempo = final_tempo * 6;

console.log("Tempo: " + final_tempo);
// final_tempo will be 126