iOS-百度语音识别

背景：

总监看我太无聊了，讯飞语音识别搞完了，就让我去搞一下百度语音识别。然后比较两种语音识别的准确率，哪种更高点。

个人认为吧，都差不多。百度稍微会好那么一点点。

不过，值得一提的是：虽然讯飞和百度的语音识别都是免费的，但是百度每天免费试用次数是20000次，而讯飞的每天只有500次。其他语音产品，也都是百度的免费次数多。讯飞和百度都有个性化词库，所谓个性化词库就是你说的语句，如果话中的词语包含在个性化词库中，那么会优先识别个性化词库的词。但是，百度除了有个性化词库之外，还有个语义解析，就是你设置好问答，然后你问上一句，他会自动答出下一句。好不好玩？讯飞不知道有没有，我没看到。

---

百度语音识别：

当然是要去百度语音开放平台注册应用啦，如果还没注册的，包括如何引入库，可以参考：iOS-百度语音合成的使用

//
//  ViewController.m
//  BirdBtn
//
//  Created by HZhenF on 2017/9/25.
//  Copyright © 2017年 GZHYTechnology. All rights reserved.
//

#import "ViewController.h"
#import "BDSEventManager.h"
#import "BDSASRDefines.h"
#import "BDSASRParameters.h"

#define ScreenW [UIScreen mainScreen].bounds.size.width
#define ScreenH [UIScreen mainScreen].bounds.size.height

const NSString* API_KEY = @"BCZkKhjhlFi10eGshX7NNEfY";
const NSString* SECRET_KEY = @"07b9795bbe95fa71a710c2179d80fc0b";
const NSString* APP_ID = @"10177498";


@interface ViewController ()<BDSClientASRDelegate>

@property(nonatomic,strong) BDSEventManager *asrManager;

@property(nonatomic,assign) BOOL isStartRecord;

@end

@implementation ViewController

- (void)viewDidLoad {
    [super viewDidLoad];


    CGFloat btn3W = 127;
    CGFloat btn3H = 93;
    CGFloat btn3X = 10;
    CGFloat btn3Y = CGRectGetMaxY(btn2.frame) + 10;
    UIButton *btn3 = [[UIButton alloc] initWithFrame:CGRectMake(btn3X, btn3Y, btn3W, btn3H)];
    [btn3 setImage:[UIImage imageNamed:@"icon_logo_L"] forState:UIControlStateNormal];

    UILongPressGestureRecognizer *longPress = [[UILongPressGestureRecognizer alloc] initWithTarget:self action:@selector(longPressAction:)];
    [btn3 addGestureRecognizer:longPress];
    [self.view addSubview:btn3];



   //配置百度语音识别
    [self configureBDVoice];
}

-(void)configureBDVoice
{
    //创建相关接口对象
    self.asrManager = [BDSEventManager createEventManagerWithName:BDS_ASR_NAME];
    //设置代理
    [self.asrManager setDelegate:self];


    //配置参数
    //1.设置DEBUG_LOG的级别
    [self.asrManager setParameter:@(EVRDebugLogLevelTrace) forKey:BDS_ASR_DEBUG_LOG_LEVEL];
    //2.配置API_KEY 和 SECRET_KEY 和 APP_ID
    [self.asrManager setParameter:@[API_KEY, SECRET_KEY] forKey:BDS_ASR_API_SECRET_KEYS];
    [self.asrManager setParameter:APP_ID forKey:BDS_ASR_OFFLINE_APP_CODE];
    //3.配置端点检测
    NSString *modelVAD_filepath = [[NSBundle mainBundle] pathForResource:@"bds_easr_basic_model" ofType:@"dat"];
    NSLog(@"modelVAD_filepath = %@",modelVAD_filepath);
    [self.asrManager setParameter:modelVAD_filepath forKey:BDS_ASR_MODEL_VAD_DAT_FILE];
    [self.asrManager setParameter:@(YES) forKey:BDS_ASR_ENABLE_MODEL_VAD];
    //4.开启语义理解
    [self.asrManager setParameter:@(YES) forKey:BDS_ASR_ENABLE_NLU];
    [self.asrManager setParameter:@"15361" forKey:BDS_ASR_PRODUCT_ID];

    //长按设置
    [self.asrManager setParameter:@(YES) forKey:BDS_ASR_VAD_ENABLE_LONG_PRESS];
    [self.asrManager setParameter:@(NO) forKey:BDS_ASR_ENABLE_LONG_SPEECH];
    [self.asrManager setParameter:@(NO) forKey:BDS_ASR_NEED_CACHE_AUDIO];
    [self.asrManager setParameter:@"" forKey:BDS_ASR_OFFLINE_ENGINE_TRIGGERED_WAKEUP_WORD];

    [self.asrManager setParameter:nil forKey:BDS_ASR_AUDIO_FILE_PATH];
    [self.asrManager setParameter:nil forKey:BDS_ASR_AUDIO_INPUT_STREAM];

    self.isStartRecord = YES;

}

-(void)longPressAction:(UILongPressGestureRecognizer *)longPress
{
//    NSLog(@"longPressAction");
    if (self.isStartRecord) {
        //发送指令:启动识别
        [self.asrManager sendCommand:BDS_ASR_CMD_START];
        self.isStartRecord = NO;
    }
    if (longPress.state == UIGestureRecognizerStateEnded) {
        self.isStartRecord = YES;
        //结束发送指令
        [self.asrManager sendCommand:BDS_ASR_CMD_STOP];
    }

}


-(void)VoiceRecognitionClientWorkStatus:(int)workStatus obj:(id)aObj
{
    switch (workStatus) {
        case EVoiceRecognitionClientWorkStatusNewRecordData: {
            NSLog(@"EVoiceRecognitionClientWorkStatusNewRecordData");
            break;
        }

        case EVoiceRecognitionClientWorkStatusStartWorkIng: {
            NSLog(@"EVoiceRecognitionClientWorkStatusStartWorkIng");
            break;
        }
        case EVoiceRecognitionClientWorkStatusStart: {
           NSLog(@"EVoiceRecognitionClientWorkStatusStart");
            break;
        }
        case EVoiceRecognitionClientWorkStatusEnd: {
            NSLog(@"EVoiceRecognitionClientWorkStatusEnd");
            break;
        }
        case EVoiceRecognitionClientWorkStatusFlushData: {
            NSLog(@"EVoiceRecognitionClientWorkStatusFlushData");
            break;
        }
        case EVoiceRecognitionClientWorkStatusFinish: {
            NSLog(@"EVoiceRecognitionClientWorkStatusFinish");

            NSLog(@"aObj = %@",aObj);

            NSArray *contentArr = aObj[@"results_recognition"];
            NSString *contentStr = contentArr[0];

            NSLog(@"contentStr = %@",contentStr);
            break;
        }
        case EVoiceRecognitionClientWorkStatusMeterLevel: {
            NSLog(@"EVoiceRecognitionClientWorkStatusMeterLevel");
            break;
        }
        case EVoiceRecognitionClientWorkStatusCancel: {
            NSLog(@"EVoiceRecognitionClientWorkStatusCancel");
            break;
        }
        case EVoiceRecognitionClientWorkStatusError: {
            NSLog(@"EVoiceRecognitionClientWorkStatusError");
            break;
        }
        case EVoiceRecognitionClientWorkStatusLoaded: {
            NSLog(@"EVoiceRecognitionClientWorkStatusLoaded");
            break;
        }
        case EVoiceRecognitionClientWorkStatusUnLoaded: {
            NSLog(@"EVoiceRecognitionClientWorkStatusUnLoaded");
            break;
        }
        case EVoiceRecognitionClientWorkStatusChunkThirdData: {
            NSLog(@"EVoiceRecognitionClientWorkStatusChunkThirdData");
            break;
        }
        case EVoiceRecognitionClientWorkStatusChunkNlu: {
            NSLog(@"EVoiceRecognitionClientWorkStatusChunkNlu");
            break;
        }
        case EVoiceRecognitionClientWorkStatusChunkEnd: {
           NSLog(@"EVoiceRecognitionClientWorkStatusChunkEnd");
            break;
        }
        case EVoiceRecognitionClientWorkStatusFeedback: {
           NSLog(@"EVoiceRecognitionClientWorkStatusFeedback");
            break;
        }
        case EVoiceRecognitionClientWorkStatusRecorderEnd: {
           NSLog(@"EVoiceRecognitionClientWorkStatusRecorderEnd");
             break;
        }
        case EVoiceRecognitionClientWorkStatusLongSpeechEnd: {
           NSLog(@"EVoiceRecognitionClientWorkStatusLongSpeechEnd");
            break;
        }
        default:
            break;
    }
}




@end

---

个性化词库和语义解析

里面有官方的格式，按照官方格式来弄就行。
弄完上传就好，个性化词库要等一个小时才能生效(讯飞、百度都是一个小时)。语义解析要等两个小时才能生效。