Microsoft Cognitive Speech Непрерывное распознавание с использованием Swift

Question

Я пытаюсь использовать Microsoft Cognitive Services Speech to Text SDK для iOS. До сих пор мне удавалось заставить функцию признания работать без проблем, но примеры для Swift редки, и я не могу показаться, что непрерывное распознавание работает со Swift. Пожалуйста, имейте в виду, что я обычно работаю с React Native / JS и у меня очень базовый c gr asp Swift в целом. При этом любая помощь будет принята с благодарностью. Это то, что я до сих пор:

func recognizeFromMic() {

    var speechConfig: SPXSpeechConfiguration?
    do {
        try speechConfig = SPXSpeechConfiguration(subscription: key, region: region)
    } catch {
        print("error \(error) happened")
        speechConfig = nil
    }
    speechConfig?.speechRecognitionLanguage = "en-US"

    let audioConfig = SPXAudioConfiguration(microphone: nil)

    let reco = try! SPXSpeechRecognizer(speechConfiguration: speechConfig!, audioConfiguration: audioConfig!)

    reco.addRecognizingEventHandler() {reco, evt in
        print("Received intermediate result event. SessionId: \(evt.sessionId), Recognition Result: \(evt.result.text ?? "(no result)"), Status: \(evt.result.reason), Offset: \(evt.result.offset) Duration: \(evt.result.duration), ResultID: \(evt.result.resultId)")
    }

    reco.addRecognizedEventHandler() {reco, evt in
        print("Received final result event. SessionId: \(evt.sessionId), Recognition Result: \(evt.result.text ?? "(no result)"), Status: \(evt.result.reason), Offset: \(evt.result.offset) Duration: \(evt.result.duration), ResultID: \(evt.result.resultId)")
    }

    reco.addSessionStoppedEventHandler() {reco, evt in
        print("Received session stopped event. SessionId: \(evt.sessionId)")
    }

    print("Listening...")

    do {

        try reco.startContinuousRecognition()

    } catch {

        print("error \(error) happened")

    }
}

При запуске вышеуказанной функции я получаю следующую ошибку:

Error Domain=SPXErrorDomain Code=21 "Error: Exception with an error code: 0x15 (SPXERR_MIC_ERROR)

Answer 1

Так что после огромной борьбы у меня наконец-то появилась рабочая функция. Microsoft определенно должна улучшить свои примеры. Для тех из вас, кто ищет одинаковые ответы, вот что я в итоге сделал. Надеюсь, это кому-нибудь поможет:)

var audioEngine = AVAudioEngine()
var audioData = Data()
var inputNode: AVAudioInputNode?
var reco: SPXSpeechRecognizer?;
var stream = SPXPullAudioInputStream()

func recognizeFromMic() {

    var bytesPerFrame = UInt32()
    var stream: SPXPushAudioInputStream?

    func toData(PCMBuffer: AVAudioPCMBuffer) -> Data {
        let channelCount = 1  // given PCMBuffer channel count is 1
        let channels = UnsafeBufferPointer(start: PCMBuffer.int16ChannelData, count: channelCount)
        let ch0Data = NSData(bytes: channels[0], length:Int(PCMBuffer.frameCapacity * PCMBuffer.format.streamDescription.pointee.mBytesPerFrame)) as Data
        return ch0Data
    }

    guard let outputFormat = AVAudioFormat(commonFormat: .pcmFormatInt16, sampleRate: 16000, channels: AVAudioChannelCount(1), interleaved: false) else {
        return
    }

    let audioSession = AVAudioSession.sharedInstance()
    do {
        try audioSession.setCategory(.record, mode: .measurement, options: .duckOthers)
        try audioSession.setActive(true, options: .notifyOthersOnDeactivation)
    } catch {

    }
    inputNode = self.audioEngine.inputNode
    let inputFormat = inputNode!.outputFormat(forBus: 0)
    let converter = AVAudioConverter(from: inputFormat, to: outputFormat)
    inputNode!.installTap(onBus: AVAudioNodeBus(0), bufferSize: AVAudioFrameCount(32000), format: self.audioEngine.inputNode.outputFormat(forBus: 0)) { (pcmBuffer, time) in

        guard let outputBuffer = AVAudioPCMBuffer(pcmFormat: outputFormat, frameCapacity:pcmBuffer.frameCapacity) else {
            return
        }
        outputBuffer.frameLength = pcmBuffer.frameLength
        let inputBlock: AVAudioConverterInputBlock = { inNumPackets, outStatus in
            outStatus.pointee = AVAudioConverterInputStatus.haveData
            return pcmBuffer
        }
        var  error:NSError? = nil
        converter?.convert(to: outputBuffer, error: &error, withInputFrom: inputBlock)
        self.audioData.append(toData(PCMBuffer: outputBuffer))
        stream?.write(self.audioData)

    }
    self.audioEngine.prepare()
    do {
        try self.audioEngine.start()
    } catch {
        print("\(error)")
    }

    //set up push stream
    let audioFormat = SPXAudioStreamFormat.init(usingPCMWithSampleRate: 16000, bitsPerSample: 16, channels: 1)
    stream = SPXPushAudioInputStream(audioFormat: audioFormat!)

    var speechConfig: SPXSpeechConfiguration?
    do {
        try speechConfig = SPXSpeechConfiguration(subscription: key, region: region)
        speechConfig!.enableDictation();
    } catch {
        print("error \(error) happened")
        speechConfig = nil
    }
    speechConfig?.speechRecognitionLanguage = "en-US"

    let audioConfig = SPXAudioConfiguration(streamInput: stream!)

    reco = try! SPXSpeechRecognizer(speechConfiguration: speechConfig!, audioConfiguration: audioConfig!)

    reco!.addRecognizingEventHandler() {reco, evt in
        print("Received intermediate result event. SessionId: \(evt.sessionId), Recognition Result: \(evt.result.text ?? "(no result)"), Status: \(evt.result.reason), Offset: \(evt.result.offset) Duration: \(evt.result.duration), ResultID: \(evt.result.resultId)")
    }

    reco!.addRecognizedEventHandler() {reco, evt in
        print("Received final result event. SessionId: \(evt.sessionId), Recognition Result: \(evt.result.text ?? "(no result)"), Status: \(evt.result.reason), Offset: \(evt.result.offset) Duration: \(evt.result.duration), ResultID: \(evt.result.resultId)")
    }

    reco!.addSessionStoppedEventHandler() {reco, evt in
        print("Received session stopped event. SessionId: \(evt.sessionId)")
    }

    print("Listening...")

    // Start recording and append recording buffer to speech recognizer

    do {
        try reco!.startContinuousRecognition()
    } catch {
        print("error \(error) happened")
    }

}