Подключение ASR в Assistant SDK Android

Обновлено 22 ноября 2023

ASR (Automatic Speech Recognition) — это распознавание речи. Вы можете интегрировать ASR в свое приложение.

Ниже описан пример интеграции ASR. Его можно найти в файле ActiveSpeechRecognizerActivity.kt, начиная со сборки 22.02.1.9010.

package ru.sberbank.sdakit.sdk.client.ext.app.presentation
 
import android.Manifest
import android.os.Bundle
import android.view.View
import androidx.appcompat.app.AppCompatActivity
import io.reactivex.Single
import io.reactivex.disposables.CompositeDisposable
import io.reactivex.disposables.Disposable
import ru.sberbank.sdakit.base.core.threading.rx.di.ThreadingRxApi
import ru.sberbank.sdakit.core.di.platform.api
import ru.sberbank.sdakit.core.platform.di.CorePlatformApi
import ru.sberbank.sdakit.core.platform.domain.permissions.ActivityBasedPermissions
import ru.sberbank.sdakit.core.platform.domain.permissions.Permissions
import ru.sberbank.sdakit.core.utils.rx.subscribeBy
import ru.sberbank.sdakit.sdk.client.ext.app.R
import ru.sberbank.sdakit.sdk.client.ext.app.databinding.ActivityActiveSpeechRecognizerBinding
import ru.sberbank.sdakit.voice.MusicRecognizer
import ru.sberbank.sdakit.voice.VoiceRecognizer
import ru.sberbank.sdakit.voice.di.VoiceRecognitionComponent
import ru.sberbank.sdakit.vps.config.ClientInfo
import ru.sberbank.sdakit.vps.config.ClientInfoFactory
import ru.sberbank.sdakit.vps.config.DefaultVpsTokenInvalidator
import ru.sberbank.sdakit.vps.config.NoopUfsMetaInfoProvider
import ru.sberbank.sdakit.vps.config.StreamingConfig
import ru.sberbank.sdakit.vps.config.TokenValue
import ru.sberbank.sdakit.vps.config.VPSClientConfig
import ru.sberbank.sdakit.vps.config.VPSTokenProvider
import ru.sberbank.sdakit.vps.config.VpsTokenMode
import ru.sberbank.sdakit.vps.config.di.VpsConfigDependencies
 
/**
 * Экран с примером перевода речи пользователя в текст.
 */
class ActiveSpeechRecognizerActivity : AppCompatActivity() {
    private val rxSchedulers by api(ThreadingRxApi::rxSchedulers)
    private val permissionsCache by api(CorePlatformApi::permissionsCache)
    private val permissionsRequestStateProvider by api(CorePlatformApi::permissionsRequestStateProvider)
 
    // Тестовый конфиг, не пригоден для продакшена. Значения для прода необходимо взять у менеджера.
    private val assistantConfig: VPSClientConfig by lazy {
        VPSClientConfig(
            serverUrl = getString(R.string.preference_assistant_vps_endpoint_uri_demo_2),
            channel = "FEBRUARY",
            clientInfo = ClientInfoFactory.create(
                surface = "DEMO_APP",
                surfaceVersion = "12.0",
                packageName = "ru.sberbank.sdakit.sdk.client.ext.app"
            ),
            streamingConfig = StreamingConfig(dubbingEnabled = false)
        )
    }
 
    private val tokenMode: VpsTokenMode = VpsTokenMode.SingleTokenProvider(object : VPSTokenProvider {
        override fun requestToken(cause: VPSTokenProvider.RequestCause): Single<TokenValue> =
            Single.just(TokenValue.esa("TEST_TOKEN"))
    })
 
    private fun vpsConfigDependencies() = object : VpsConfigDependencies {
        override val vpsClientConfig = assistantConfig
        override val vpsTokenMode = tokenMode
        override val vpsTokenInvalidator = DefaultVpsTokenInvalidator()
        override val ufsMetaInfoProvider = NoopUfsMetaInfoProvider()
    }
 
    private val voiceRecognizer: VoiceRecognizer by lazy {
        VoiceRecognitionComponent.createWithDependencies(vpsConfigDependencies()).voiceRecognizer
    }
 
    private val musicRecognizer: MusicRecognizer by lazy {
        VoiceRecognitionComponent.createWithDependencies(vpsConfigDependencies()).musicRecognizer
    }
 
    private lateinit var binding: ActivityActiveSpeechRecognizerBinding
    private lateinit var permissions: Permissions
 
    private val disposables = CompositeDisposable()
 
    private var started = false
 
    override fun onCreate(savedInstanceState: Bundle?) {
        super.onCreate(savedInstanceState)
 
        binding = ActivityActiveSpeechRecognizerBinding.inflate(layoutInflater)
        setContentView(binding.root)
 
        binding.recognizedText.text = ""
        updateButtonAppearance()
 
        binding.startStopButton.setOnClickListener { startStop() }
 
        binding.recognitionTypeVoice.isChecked = true
 
        binding.recordingProgress.visibility = View.INVISIBLE
 
        permissions = ActivityBasedPermissions(this, permissionsCache, permissionsRequestStateProvider)
    }
 
    override fun onRequestPermissionsResult(requestCode: Int, permissions: Array<String>, grantResults: IntArray) {
        super.onRequestPermissionsResult(requestCode, permissions, grantResults)
        this.permissions.onRequestPermissionsResult(requestCode, permissions, grantResults)
    }
 
    override fun onStop() {
        super.onStop()
        stop()
    }
 
    private fun startStop() = when (started) {
        true -> stop()
        false -> start()
    }
 
    private fun startRecognition(): Disposable =
        when (binding.recognitionType.checkedRadioButtonId) {
            R.id.recognition_type_voice -> startVoiceRecognition()
            R.id.recognition_type_music -> startMusicRecognition()
            else -> startVoiceRecognition()
        }
 
    private fun startVoiceRecognition(): Disposable =
        voiceRecognizer.startRecognition()
            .observeOn(rxSchedulers.ui())
            .subscribeBy(
                onNext = {
                    binding.recognizedText.text = it
                },
                onComplete = {
                    stop()
                },
                onError = {
                    stop()
                }
            )
 
    private fun startMusicRecognition(): Disposable =
        musicRecognizer.startRecognition()
            .observeOn(rxSchedulers.ui())
            .subscribeBy(
                onNext = { value ->
                    binding.recognizedText.text = if (value.isFinal) value.response else "В процессе"
                },
                onComplete = {
                    stop()
                },
                onError = {
                    stop()
                }
            )
 
    private fun start() {
        started = true
 
        binding.recognizedText.text = ""
        binding.recordingProgress.visibility = View.VISIBLE
        updateButtonAppearance()
 
        disposables.add(
            voiceRecognizer.observeAudioRecordingPermissionRequests()
                .observeOn(rxSchedulers.ui())
                .subscribeBy(
                    onNext = {
                        permissions.request(Manifest.permission.RECORD_AUDIO)
                    },
                    onError = {}
                )
        )
 
        disposables.add(startRecognition())
    }
 
    private fun stop() {
        started = false
 
        binding.recordingProgress.visibility = View.INVISIBLE
 
        updateButtonAppearance()
 
        voiceRecognizer.stopRecognition()
        musicRecognizer.stopRecognition()
        disposables.clear()
    }
 
    @Suppress("SetTextI18n")
    private fun updateButtonAppearance() = when (started) {
        true -> binding.startStopButton.text = "ОСТАНОВИТЬ"
        false -> binding.startStopButton.text = "РАСПОЗНАТЬ$recognitionTypeName"
    }
 
    private val recognitionTypeName: String
        get() = when (binding.recognitionType.checkedRadioButtonId) {
            R.id.recognition_type_voice -> " ГОЛОС"
            R.id.recognition_type_music -> " МУЗЫКУ"
            else -> ""
        }
}