Source code
Revision control
Copy as Markdown
Other Tools
Test Info: Warnings
- This test gets skipped with pattern: os == 'linux' && os_version == '22.04' && arch == 'x86_64' && display == 'wayland' OR verify
- Manifest: dom/media/webspeech/recognition/test/mochitest.toml
<!DOCTYPE HTML>
<html>
<head>
<meta charset="utf-8">
<title>Test SpeechRecognition with a one-word utterance</title>
<script src="/tests/SimpleTest/SimpleTest.js"></script>
<script src="head.js"></script>
<link rel="stylesheet" type="text/css" href="/tests/SimpleTest/test.css"/>
</head>
<body>
<div id="content"></div>
<div id="results">
<label>Recognized:</label>
<div class="recognized" id="rec-text">waiting...</div>
</div>
<pre id="test">
<script>
"use strict";
// Plays a one-word clip, then feeds 3s of silence, and checks the word is
// delivered once, before stop(). Uses the default English model, which marks
// utterance ends itself.
const SERVER_PORT = 8766;
const LANG = "en-US";
const WORD = "dog";
// Silence after the word, where a repeat would show up.
const LISTEN_AFTER_MS = 3000;
SimpleTest.requestCompleteLog();
SimpleTest.requestFlakyTimeout(
"createResumedAudioContext() is bounded by setTimeout, and the silence after " +
"the word is waited out with one");
const T0 = performance.now();
const ms = () => "+" + Math.round(performance.now() - T0) + "ms";
function diag(msg) { info("[diag " + ms() + "] " + msg); }
function words(text) {
return text.toLowerCase().replace(/[^\p{L}\p{N}\s']/gu, " ").split(/\s+/).filter(Boolean);
}
add_setup(async function setup() {
await SpecialPowers.pushPrefEnv({
set: [
["media.webspeech.recognition.enable", true],
["media.webspeech.recognition.model-download.prompt.testing", true],
["media.navigator.permission.disabled", true],
],
});
});
add_task(async function one_word_one_result() {
const installed = await ensureModelInstalled([LANG]);
ok(installed, "Model installed successfully");
if (!installed) {
return;
}
const ctx = await createResumedAudioContext();
// Keeps the graph running past the end of the clip, so we feed silence.
const kick = ctx.createBufferSource();
kick.buffer = ctx.createBuffer(1, 2048, ctx.sampleRate);
kick.loop = true;
kick.connect(ctx.destination);
kick.start(0);
const audio = await loadTestAudio("single_word_dog.opus", a => {
a.id = "audioElement";
a.controls = true;
});
const src = ctx.createMediaElementSource(audio);
const dst = ctx.createMediaStreamDestination();
src.connect(dst);
src.connect(ctx.destination);
const track = dst.stream.getAudioTracks()[0];
ok(track, "Got audio track");
const sr = new SpeechRecognition();
sr.continuous = true;
sr.lang = LANG;
const finals = [];
sr.onresult = e => {
for (let i = e.resultIndex; i < e.results.length; i++) {
if (e.results[i].isFinal) {
const text = e.results[i][0].transcript;
finals.push({ text, at: performance.now() });
diag(`final result #${finals.length}: "${text}"`);
}
}
document.getElementById("rec-text").textContent =
finals.map(f => f.text).join(" | ");
};
const ended = new Promise((resolve, reject) => {
sr.onend = resolve;
sr.onerror = e => reject(new Error(e.error));
});
sr.start(track);
await audio.play();
await new Promise(r => { audio.onended = r; });
diag(`word over, listening to silence for ${LISTEN_AFTER_MS}ms`);
await new Promise(r => setTimeout(r, LISTEN_AFTER_MS));
const stoppedAt = performance.now();
sr.stop();
await ended;
audio.pause();
await ctx.close();
const listening = finals.filter(f => f.at < stoppedAt);
const transcript = finals.map(f => f.text).join(" | ");
is(listening.length, 1,
`One word is one final result, delivered while the session is still ` +
`listening. Got ${listening.length} of ${finals.length}: "${transcript}"`);
if (!listening.length) {
return;
}
is(words(listening[0].text).join(" "), WORD,
`The result is the word that was said. Got: "${listening[0].text}"`);
is(finals.length, 1,
`The word is not delivered again over the silence that follows it, nor ` +
`by the end-of-stream flush. Got ${finals.length}: "${transcript}"`);
});
</script>
</pre>
</body>
</html>