Skip to content

suifengqjn/IOS10Speech

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

IOS10 语音识别详解

公司项目需要实现语音搜索,正好记录一下这个iOS10新出的API。

iOS10是一个变化比价大的版本,开放了很多借口,这样也更方便开发者自定义各种功能。本文主要讲解一下新增的Speech框架,有了这个框架,我们想要为自己的app增加语音识别功能,不要依赖第三方的服务,几十行代码就可以轻松搞定。demo地址在文章末尾。

一:基本配置

  • Xcode8,iOS10系统真机
  • 导入头文件:OC #import<Speech/Speech.h> swift import Speech
  • 配置info.plist文件:配置两个权限,语音识别和麦克风
<key>NSMicrophoneUsageDescription</key>
    <string>Your microphone will be used to record your speech when you press the "Start Recording" button.</string>
    
    <key>NSSpeechRecognitionUsageDescription</key>
    <string>Speech recognition will be used to determine which words you speak into this device's microphone.</string>
  

二:用到的几个类

AVAudioEngine 语音引擎,负责提供语音输入 SFSpeechAudioBufferRecognitionRequest 处理语音识别请求 SFSpeechRecognizer 语音识别器 SFSpeechRecognitionTask 输出语音识别对象的结果 NSLocale 语言类型
语音识别一共就用到了这几个类,整体的流程也容易理解,语音识别器通过语音引擎,处理语音识别请求,把结果交给SFSpeechRecognitionTask处理,最后输出文字。 SFSpeechRecognizer 自身有几个代理方法,实际上,如果只是将语音转化成文字,是不需要这几个代理方法的。

//当开始检测音频源中的语音时首先调用此方法
-(void)speechRecognitionDidDetectSpeech:(SFSpeechRecognitionTask *)task
{
  
}
//当识别出一条可用的信息后 会调用
/*需要注意,apple的语音识别服务会根据提供的音频源识别出多个可能的结果 每有一条结果可用 都会调用此方法 */
-(void)speechRecognitionTask:(SFSpeechRecognitionTask *)task didHypothesizeTranscription:(SFTranscription *)transcription
{
   
}
//当识别完成所有可用的结果后调用
- (void)speechRecognitionTask:(SFSpeechRecognitionTask *)task didFinishRecognition:(SFSpeechRecognitionResult *)recognitionResult
{
    
}
//当不再接受音频输入时调用 即开始处理语音识别任务时调用
- (void)speechRecognitionTaskFinishedReadingAudio:(SFSpeechRecognitionTask *)task
{
    
}
//当语音识别任务被取消时调用
- (void)speechRecognitionTaskWasCancelled:(SFSpeechRecognitionTask *)task
{
    
}
//语音识别任务完成时被调用
- (void)speechRecognitionTask:(SFSpeechRecognitionTask *)task didFinishSuccessfully:(BOOL)successfully
{

}

三:重点代码

有两点需要注意:

  • 语音识别会很耗电以及会使用很多数据
  • 语音识别一次只持续大概一分钟的时间

我先定义了这几个属性

@property (nonatomic, strong) AVAudioEngine         *audioEngine;
@property (nonatomic, strong) SFSpeechRecognizer    *speechRecognizer;
@property (nonatomic, strong) SFSpeechAudioBufferRecognitionRequest     *recognitionRequest;
@property (nonatomic, strong) SFSpeechRecognitionTask   *recognitionTask;
@property (nonatomic, strong) NSLocale                  *locale;
  1. 语音权限的判断
[SFSpeechRecognizer requestAuthorization:^(SFSpeechRecognizerAuthorizationStatus status) {
        BOOL isAuthorized = NO;
       switch (status) {
                //结果未知 用户尚未进行选择
            case SFSpeechRecognizerAuthorizationStatusNotDetermined:
                isAuthorized = NO;
                break;
                //用户拒绝授权语音识别
            case SFSpeechRecognizerAuthorizationStatusDenied:
                isAuthorized = NO;
                break;
                //设备不支持语音识别功能
            case SFSpeechRecognizerAuthorizationStatusRestricted:
                isAuthorized = NO;
                break;
                //用户授权语音识别
            case SFSpeechRecognizerAuthorizationStatusAuthorized:
                isAuthorized = YES;
                
                break;
                
            default:
                break;
        }
        
        if (callback) {
            callback(isAuthorized, status);
        }
    }];
  1. 将语音引擎得到的语音数据添加到语音识别的请求中,这个过程也就是开始录音后的流程
AVAudioFormat *recordingFormat = [[self.audioEngine inputNode] outputFormatForBus:0];
    [[self.audioEngine inputNode] installTapOnBus:0 bufferSize:1024 format:recordingFormat block:^(AVAudioPCMBuffer * _Nonnull buffer, AVAudioTime * _Nonnull when) {
        [self.recognitionRequest appendAudioPCMBuffer:buffer];
    }];
  1. SFSpeechRecognitionTask 把上一过程中得到的语音请求转化成文字,这个过程是试试进行的。
self.recognitionTask = [self.speechRecognizer recognitionTaskWithRequest:self.recognitionRequest resultHandler:^(SFSpeechRecognitionResult * _Nullable result, NSError * _Nullable error) {
        BOOL isFinal = NO;
        NSString *bestResult = [[result bestTranscription] formattedString];
        isFinal = result.isFinal;
        if (error || isFinal) {
            [self endTask];
            if (self.delegate && [self.delegate respondsToSelector:@selector(recognizeFail:)]) {
                [self.delegate recognizeFail:error];
            }
        } else {
            if (self.delegate && [self.delegate respondsToSelector:@selector(recognizeSuccess:)]) {
                [self.delegate recognizeSuccess:bestResult];
            }
        }
    }];

四:提取录音文件中的文字

  1. 也需要先获取用户的授权,授权代码与上面一致。
  2. 对文件的处理相对较为简单
    //初始化一个识别器
    SFSpeechRecognizer *recognizer = [[SFSpeechRecognizer alloc] initWithLocale:[NSLocale localeWithLocaleIdentifier:@"zh_CN"]];
    //初始化mp3的url
    NSURL *url = [[NSBundle mainBundle] URLForResource:@"test.mp3" withExtension:nil];
    //初始化一个识别的请求
    SFSpeechURLRecognitionRequest *request = [[SFSpeechURLRecognitionRequest alloc] initWithURL:url];
    //发起请求
    [recognizer recognitionTaskWithRequest:request resultHandler:^(SFSpeechRecognitionResult * _Nullable result, NSError * _Nullable error) {
        if(error != nil)
        {
            NSLog(@"识别错误:%@",error);
        }
        NSString *resultString = result.bestTranscription.formattedString;
        NSLog(@"%@",resultString);
        
    }];

个人blog:http://gcblog.github.io 简书blog:http://www.jianshu.com/users/527ecf8c8753/latest_articles

About

实时语音转文字,以及录音文件转文字

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published