* 👋 Hello! This is an ml5.js example made and shared with ❤️.

 * Learn more about the ml5.js project: https://ml5js.org/

 * ml5.js license and Code of Conduct: https://github.com/ml5js/ml5-next-gen/blob/main/LICENSE.md

 *

 * This example demonstrates Sound classification using SpeechCommands18w

 */

​

// Initialize a sound classifier method with SpeechCommands18w model. A callback needs to be passed.

let classifier;

​

// Array containing the 18 words of SpeechCommands18w

let words = [

  "zero",

  "one",

  "two",

  "three",

  "four",

  "five",

  "six",

  "seven",

  "eight",

  "nine",

  "up",

  "down",

  "left",

  "right",

  "go",

  "stop",

  "yes",

  "no",

];

​

// Variable for displaying the results on the canvas

let predictedWord = "";

​

function preload() {

  // Options for the SpeechCommands18w model, the default probabilityThreshold is 0

  let options = { probabilityThreshold: 0.7 };

  // Load SpeechCommands18w sound classifier model

  classifier = ml5.soundClassifier("SpeechCommands18w", options);

}

​

function setup() {

  createCanvas(650, 450);

  // Classify the sound from microphone in real time

  classifier.classifyStart(gotResult);

}

​

function draw() {

  background(250);

  // Call function for displaying background words

  displayWords();

​

  // Once the model outputs results start displaying the predicted word on the canvas

  if (predictedWord !== "") {

    fill(211, 107, 255);

    textAlign(CENTER, CENTER);

    textSize(64);

    text(predictedWord, width / 2, 90);

  }

}

​

// Function to display the 18 words on the canvas

function displayWords() {

  textAlign(CENTER, CENTER);

  textSize(32);

  fill(96);

  text("Say one of these words!", width / 2, 40);

​

  let x = 125;

  let y = 150;

  // Words appear in 3 columns of 6 rows

  for (let i = 0; i < words.length; i++) {

    fill(158);

    text(words[i], x, y);

    y += 50;

    if ((i + 1) % 6 === 0) {

      x += 200;

      y = 150;

    }

  }

}

​

// A function to run when we get any errors and the results

function gotResult(results) {

  // The results are in an array ordered by confidence

  console.log(results);

  // Load the first label to the text variable displayed on the canvas

  predictedWord = results[0].label;

}

​