這篇日志記錄自己在學(xué)習(xí)使用蘋果原生框架語音識(shí)別庫Speech Framework時(shí)的總結(jié)和示例代碼触趴。看了一遍官方文檔，把該框架中的相關(guān)類和方法了解了一遍殖演，然后總結(jié)了一張XMind結(jié)構(gòu)圖熄浓。

前提：需要Xcode 8 以上和一個(gè)運(yùn)行iOS10以上系統(tǒng)的iOS設(shè)備.

Speech Framework中的類和方法概念

Paste_Image.png

Note： 因?yàn)樯婕暗綑?quán)限問題情臭，需要在info.plist文件中添加兩個(gè)key。分別是Privacy - Microphone Usage Description(麥克風(fēng)權(quán)限)和 Privacy - Speech Recognition Usage Description(語音識(shí)別權(quán)限)

Swift代碼

import UIKit
import Speech

class ViewController: UIViewController {

  @IBOutlet weak var textView: UITextView!
  @IBOutlet weak var microphoneButton: UIButton!

  /// 語音識(shí)別操作類對(duì)象
  private let speechRecognizer = SFSpeechRecognizer()

  /// 處理語音識(shí)別請(qǐng)求赌蔑，給語音識(shí)別提供語音輸入
  private var recognitionRequest: SFSpeechAudioBufferRecognitionRequest?

  /// 告訴用戶語音識(shí)別對(duì)象的結(jié)果俯在。擁有這個(gè)對(duì)象很方便因?yàn)槟憧梢?用它刪除或中斷任務(wù)
  private var recognitionTask: SFSpeechRecognitionTask?

  /// 語音引擎。負(fù)責(zé)提供語音輸入
  private let audioEngine = AVAudioEngine()

  override func viewDidLoad() {
      super.viewDidLoad()
      // Do any additional setup after loading the view, typically from a nib.
      microphoneButton.isEnabled = false
    
      speechRecognizer?.delegate = self

      /// 申請(qǐng)用戶語音識(shí)別權(quán)限
      SFSpeechRecognizer.requestAuthorization { (authStatus) in

          var isButtonEnabled = false

          switch authStatus {
          case .authorized: // 用戶授權(quán)語音識(shí)別
              isButtonEnabled = true

          case .denied: // 用戶拒絕授權(quán)語音識(shí)別
              isButtonEnabled = false
              print("User denied access to speech recognition")

          case .restricted: // 設(shè)備不支持語音識(shí)別功能
              isButtonEnabled = false
              print("Speech recognition restricted on this device")

          case .notDetermined: // 結(jié)果未知 用戶尚未進(jìn)行選擇
              isButtonEnabled = false
              print("Speech recognition not yet authorized")
          }

          OperationQueue.main.addOperation {
              self.microphoneButton.isEnabled = isButtonEnabled
          }
      }
  }

  @IBAction func microphoneButtonClick(_ sender: UIButton) {
      if audioEngine.isRunning {
          audioEngine.stop()
          recognitionRequest?.endAudio()
          microphoneButton.isEnabled = false
          microphoneButton.setTitle("Start Recording", for: .normal)
      } else {
          startRecording()
          microphoneButton.setTitle("Stop Recording", for: .normal)
      }
  }

  func startRecording() {
      if recognitionTask != nil { /// 檢查recognitionTask是否在運(yùn)行娃惯，如果在就取消任務(wù)和識(shí)別
          recognitionTask?.cancel()
          recognitionTask = nil
      }

      let audioSession = AVAudioSession.sharedInstance() /// 記錄語音做準(zhǔn)備
      do {
          try audioSession.setCategory(AVAudioSessionCategoryRecord)
          try audioSession.setMode(AVAudioSessionModeMeasurement)
          try audioSession.setActive(true, with: .notifyOthersOnDeactivation)
      } catch {
          print("audioSession properties weren't set because of an error.")
      }

      /// 實(shí)例化recognitionRequest 利用它把語音數(shù)據(jù)傳到蘋果后臺(tái)
      recognitionRequest = SFSpeechAudioBufferRecognitionRequest()

      /// 檢查audioEngine(你的設(shè)備)是否有做錄音功能作為語音輸入
      guard let inputNode = audioEngine.inputNode else {
          fatalError("Audio engine has no input node")
      }

      /// 檢查recognitionRequest對(duì)象是否被實(shí)例化或不是nil
      guard let recognitionRequest = recognitionRequest else {
          fatalError("Unable to create an SFSpeechAudioBufferRecongitionRequest object")
    }

      /// 當(dāng)用戶說話的時(shí)候讓recognitionRequest報(bào)告語音識(shí)別的部分結(jié)果
      recognitionRequest.shouldReportPartialResults = true

      /// 開啟語音識(shí)別, 回調(diào)每次都會(huì)在識(shí)別引擎收到輸入的時(shí)候跷乐，完善了當(dāng)前識(shí)別的信息時(shí)候，或者被刪除或者停止的時(shí)候被調(diào)用趾浅，最后會(huì)返回一個(gè)最終的文本
      recognitionTask = speechRecognizer?.recognitionTask(with: recognitionRequest, resultHandler: { (result, error) in
          var isFinal = false // 定義一個(gè)布爾值決定識(shí)別是否已經(jīng)結(jié)束

          /// 如果結(jié)果result不是nil愕提，把textView.text的值設(shè)置為我們的最優(yōu)文本。如果結(jié)果是最終結(jié)果皿哨，設(shè)置isFinal為true
          if result != nil {
              self.textView.text = result?.bestTranscription.formattedString
              isFinal = (result?.isFinal)!
          }

          /// 如果沒有錯(cuò)誤或者結(jié)果是最終結(jié)果浅侨，停止audioEngine(語音輸入)并且停止recognitionRequest和recognitionTask
          if error != nil || isFinal {
              self.audioEngine.stop()
              inputNode.removeTap(onBus: 0)

              self.recognitionRequest = nil
              self.recognitionTask = nil

              self.microphoneButton.isEnabled = true
          }
      })

      /// 向recognitionRequest增加一個(gè)語音輸入。注意在開始了recognitionTask之后增加語音輸入是OK的证膨。SpeechFramework會(huì)在語音輸入被加入的同時(shí)就開始進(jìn)行解析識(shí)別
      let recordingFormat = inputNode.outputFormat(forBus: 0)
      inputNode.installTap(onBus: 0, bufferSize: 1024, format: recordingFormat) { (buffer, when) in
          self.recognitionRequest?.append(buffer)
      }

      /// 準(zhǔn)備并且開始audioEngine
      audioEngine.prepare()

      do {
          try audioEngine.start()
      } catch {
          print("audioEngine couldn't start because of an error")
     }

      textView.text = "Say something, I'm listening!"
  }
}

extension ViewController: SFSpeechRecognizerDelegate {
  /// 可用性狀態(tài)改變時(shí)被調(diào)用
  func speechRecognizer(_ speechRecognizer: SFSpeechRecognizer, availabilityDidChange available: Bool) {
      if available {
          microphoneButton.isEnabled = true
      } else {
          microphoneButton.isEnabled = false
      }
  }
}  

Objective-C 代碼

#import <Speech/Speech.h>

@interface ViewController ()<SFSpeechRecognizerDelegate>

@property (nonatomic, strong) SFSpeechRecognizer *speechRecognizer;

@property (nonatomic, strong) SFSpeechRecognitionTask *recognitionTask;

@property (nonatomic, strong) SFSpeechAudioBufferRecognitionRequest *recognitionRequest;

/// 音頻引擎
@property (nonatomic, strong) AVAudioEngine *audioEngine;

@property (weak, nonatomic) IBOutlet UITextView *textView;
@property (weak, nonatomic) IBOutlet UIButton *microphoneBtn;

@end

@implementation ViewController

- (void)dealloc {
    [self.recognitionTask cancel];
    self.recognitionTask = nil;
}

- (void)viewDidLoad {
    [super viewDidLoad];
    // Do any additional setup after loading the view, typically from a nib.
    self.view.backgroundColor = [UIColor whiteColor];

    NSLog(@"supportedLocales: %@", [SFSpeechRecognizer supportedLocales]);

    self.microphoneBtn.enabled = NO;

    /// 創(chuàng)建語音識(shí)別器對(duì)象并設(shè)置代理
    self.speechRecognizer = [[SFSpeechRecognizer alloc] init];

    self.speechRecognizer.delegate = self;

    /// 請(qǐng)求用戶授權(quán)
    [SFSpeechRecognizer requestAuthorization:^(SFSpeechRecognizerAuthorizationStatus status) {

    BOOL isButtonEnabled = NO;

    switch (status) {
        case SFSpeechRecognizerAuthorizationStatusNotDetermined:
            isButtonEnabled = NO;
            NSLog(@"SFSpeechRecognizerAuthorizationStatusNotDetermined");
            break;
        case SFSpeechRecognizerAuthorizationStatusDenied:
            isButtonEnabled = NO;
            NSLog(@"SFSpeechRecognizerAuthorizationStatusDenied");
            break;
        case SFSpeechRecognizerAuthorizationStatusRestricted:
            isButtonEnabled = NO;
            NSLog(@"SFSpeechRecognizerAuthorizationStatusRestricted");
            break;
        case SFSpeechRecognizerAuthorizationStatusAuthorized:
            NSLog(@"SFSpeechRecognizerAuthorizationStatusAuthorized");
            isButtonEnabled = YES;
            break;
        default:
            break;
    }

    dispatch_async(dispatch_get_main_queue(), ^{
        self.microphoneBtn.enabled = isButtonEnabled;
    });
}];


    /// 創(chuàng)建音頻引擎對(duì)象
    self.audioEngine = [[AVAudioEngine alloc] init];
}

- (IBAction)microphoneBtnClick:(UIButton *)sender {
    if (self.audioEngine.isRunning) {
        [self.audioEngine stop];
        [self.recognitionRequest endAudio];
        self.microphoneBtn.enabled = NO;
        [self.microphoneBtn setTitle:@"Start Recording" forState:UIControlStateNormal];
    } else {
        [self startRecording];
        [self.microphoneBtn setTitle:@"Stop Recording" forState:UIControlStateNormal];
    }
}

#pragma mark - private method
- (void)startRecording {
    if (self.recognitionTask != nil) {
        [self.recognitionTask cancel]; // 取消當(dāng)前語音識(shí)別任務(wù)
        self.recognitionTask = nil;
    }

    AVAudioSession *audioSession = [AVAudioSession sharedInstance];
    NSError *categoryError = nil;
    if (![audioSession setCategory:AVAudioSessionCategoryRecord error:&categoryError]) {
        NSLog(@"categoryError: %@", categoryError.localizedDescription);
    }

    NSError *modeError = nil;
    if (![audioSession setMode:AVAudioSessionModeMeasurement error:&modeError]) {
        NSLog(@"modeError: %@", modeError.localizedDescription);
    }

    NSError *activeError = nil;
    if (![audioSession setActive:YES withOptions:AVAudioSessionSetActiveOptionNotifyOthersOnDeactivation error:&activeError]) {
        NSLog(@"activeError: %@", activeError.localizedDescription);
    }

    /// 實(shí)例化 通過設(shè)備麥克風(fēng)識(shí)別現(xiàn)場(chǎng)語音的請(qǐng)求 對(duì)象
    self.recognitionRequest = [[SFSpeechAudioBufferRecognitionRequest alloc] init];

    if (!self.audioEngine.inputNode) {// 系統(tǒng)輸入節(jié)點(diǎn)
        NSLog(@"Audio engine has no input node");
        return;
    }

    if (!self.recognitionRequest) {
        NSLog(@"Unable to create an SFSpeechAudioBufferRecongitionRequest object");
        return;
    }

    /// 報(bào)告每個(gè)發(fā)音的部分非精確結(jié)果
    self.recognitionRequest.shouldReportPartialResults = YES;

    /// 執(zhí)行語音識(shí)別任務(wù) 完成回調(diào)
    self.recognitionTask = [self.speechRecognizer recognitionTaskWithRequest:self.recognitionRequest resultHandler:^(SFSpeechRecognitionResult * _Nullable result, NSError * _Nullable error) {

    BOOL isFinal = NO;

    if (result) {
        self.textView.text = result.bestTranscription.formattedString;
        isFinal = result.isFinal;
    }

    if (error || isFinal) {
        [self.audioEngine stop];
        [self.audioEngine.inputNode removeTapOnBus:0];

        self.recognitionRequest = nil;
        self.recognitionTask = nil;

        self.microphoneBtn.enabled = YES;
    }
}];

    AVAudioFormat *recordingFormat = [self.audioEngine.inputNode outputFormatForBus:0];

    [self.audioEngine.inputNode installTapOnBus:0 bufferSize:1024 format:recordingFormat block:^(AVAudioPCMBuffer * _Nonnull buffer, AVAudioTime * _Nonnull when) {
        /// 將PCM格式的音頻追加到識(shí)別請(qǐng)求的結(jié)尾
        [self.recognitionRequest appendAudioPCMBuffer:buffer];
    }];

    [self.audioEngine prepare];

    NSError *startError = nil;
    if(![self.audioEngine startAndReturnError:&startError]) {
        NSLog(@"startError: %@", startError.localizedDescription);
    }

    self.textView.text = @"Say something, I'm listening";
}

#pragma mark - SFSpeechRecognizerDelegate
- (void)speechRecognizer:(SFSpeechRecognizer *)speechRecognizer availabilityDidChange:(BOOL)available {
    if (available) {
        self.microphoneBtn.enabled = YES;
    } else {
        self.microphoneBtn.enabled = NO;
    }
}

參考鏈接：
Building a Speech-to-Text App Using Speech Framework in iOS 10
SpeakToMe: Using Speech Recognition with AVAudioEngine