jiti-meet/react/features/base/devices/services/device-detect/ActiveDeviceDetector.js

// @flow
import EventEmitter from 'events';
import { ACTIVE_DEVICE_DETECTED } from './Events';
import logger from '../../logger';
import JitsiMeetJS from '../../../lib-jitsi-meet';

const JitsiTrackEvents = JitsiMeetJS.events.track;

// If after 3000 ms the detector did not find any active devices consider that there aren't any usable ones available
// i.e. audioLevel > 0.008
const DETECTION_TIMEOUT = 3000;

/**
 * Detect active input devices based on their audio levels, currently this is very simplistic. It works by simply
 * checking all monitored devices for TRACK_AUDIO_LEVEL_CHANGED if a device has a audio level > 0.008 ( 0.008 is
 * no input from the perspective of a JitsiLocalTrack ), at which point it triggers a ACTIVE_DEVICE_DETECTED event.
 * If there are no devices that meet that criteria for DETECTION_TIMEOUT an event with empty deviceLabel parameter
 * will be triggered,
 * signaling that no active device was detected.
 * TODO Potentially improve the active device detection using rnnoise VAD scoring.
 */
export class ActiveDeviceDetector extends EventEmitter {

    /**
     * Currently monitored devices.
     */
    _availableDevices: Array<Object>;

    /**
     * State flag, check if the instance was destroyed.
     */
    _destroyed: boolean = false;

    /**
     * Create active device detector.
     *
     * @param {Array<MediaDeviceInfo>} micDeviceList - Device list that is monitored inside the service.
     *
     * @returns {ActiveDeviceDetector}
     */
    static async create(micDeviceList: Array<MediaDeviceInfo>) {
        const availableDevices = [];

        try {
            for (const micDevice of micDeviceList) {
                const localTrack = await JitsiMeetJS.createLocalTracks({
                    devices: [ 'audio' ],
                    micDeviceId: micDevice.deviceId
                });

                // We provide a specific deviceId thus we expect a single JitsiLocalTrack to be returned.
                availableDevices.push(localTrack[0]);
            }

            return new ActiveDeviceDetector(availableDevices);
        } catch (error) {
            logger.error('Cleaning up remaining JitsiLocalTrack, due to ActiveDeviceDetector create fail!');

            for (const device of availableDevices) {
                device.stopStream();
            }

            throw error;
        }
    }

    /**
     * Constructor.
     *
     * @param {Array<Object>} availableDevices - Device list that is monitored inside the service.
     */
    constructor(availableDevices: Array<Object>) {
        super();

        this._availableDevices = availableDevices;

        // Setup event handlers for monitored devices.
        for (const device of this._availableDevices) {
            device.on(JitsiTrackEvents.TRACK_AUDIO_LEVEL_CHANGED, audioLevel => {
                this._handleAudioLevelEvent(device, audioLevel);
            });
        }

        // Cancel the detection in case no devices was found with audioLevel > 0 in te set timeout.
        setTimeout(this._handleDetectionTimeout.bind(this), DETECTION_TIMEOUT);
    }

    /**
     * Handle what happens if no device publishes a score in the defined time frame, i.e. Emit an event with empty
     * deviceLabel.
     *
     * @returns {void}
     */
    _handleDetectionTimeout() {
        if (!this._destroyed) {
            this.emit(ACTIVE_DEVICE_DETECTED, { deviceLabel: '',
                audioLevel: 0 });
            this.destroy();
        }
    }

    /**
     * Handles audio level event generated by JitsiLocalTracks.
     *
     * @param {Object} device - Label of the emitting track.
     * @param {number} audioLevel - Audio level generated by device.
     *
     * @returns {void}
     */
    _handleAudioLevelEvent(device, audioLevel) {
        if (!this._destroyed) {
            // This is a very naive approach but works is most, a more accurate approach would ne to use rnnoise
            // in order to limit the number of false positives.
            // The 0.008 constant is due to how LocalStatsCollector from lib-jitsi-meet publishes audio-levels, in this
            // case 0.008 denotes no input.
            // TODO potentially refactor lib-jitsi-meet to expose this constant as a function. i.e. getSilenceLevel.
            if (audioLevel > 0.008) {
                this.emit(ACTIVE_DEVICE_DETECTED, { deviceId: device.deviceId,
                    deviceLabel: device.track.label,
                    audioLevel });
                this.destroy();
            }
        }
    }

    /**
     * Destroy the ActiveDeviceDetector, clean up the currently monitored devices associated JitsiLocalTracks.
     *
     * @returns {void}.
     */
    destroy() {
        if (this._destroyed) {
            return;
        }

        for (const device of this._availableDevices) {
            device.removeAllListeners();
            device.stopStream();
        }

        this._destroyed = true;
    }
}