引言

去年在一家公司实习，接触了一点人工智能。既然接触人工智能，那么语音识别、语音合成、人脸识别等等都是必不可少的。
本来已经是离开iOS开发这条路了，还剩下毕业设计要做，所以这段时间最后一次回来接触iOS了。
以后，估计没什么机会再接触iOS，都往着机器学习、深度学习方向去了。
废话不说了，记录一下，如何接入讯飞语音。

---

步骤

• 一、官网下载SDK
  去讯飞语音开放平台下载iOS的SDK
  讯飞语音开放平台

当然，你需要创建一个应用，然后才能下载iOS的SDK，应用选择语音识别的类型。(这里只是演示语音识别类型的接入，其余语音合成、人脸识别等等自己可以试试)

• 二、工程导入相应的包和文件

这里仅演示在线语音识别，离线的请按照文档来操作
iOS讯飞操作文档

• 三、设置系统参数
  

用户隐私权限配置

iOS 10发布以来，苹果为了用户信息安全，加入隐私权限设置机制，让用户来选择是否允许。
隐私权限配置可在info.plist 新增相关privacy字段，MSC SDK中需要用到的权限主要包括麦克风权限、联系人权限和地理位置权限：

<key>NSMicrophoneUsageDescription</key>
<string></string>
<key>NSLocationUsageDescription</key>
<string></string>
<key>NSLocationAlwaysUsageDescription</key>
<string></string>
<key>NSContactsUsageDescription</key>
<string></string>

• 四、代码初始化

    //Appid是应用的身份信息，具有唯一性，初始化时必须要传入Appid。
    NSString *initString = [[NSString alloc] initWithFormat:@"appid=%@",xunfeiID];
    [IFlySpeechUtility createUtility:initString];

ID就是对应图片的ID。
建议把上面这段代码放在AppDelegate.m中

- (BOOL)application:(UIApplication *)application didFinishLaunchingWithOptions:(NSDictionary *)launchOptions

方法里面初始化。

PS：要导入头文件#import "IFlyMSC/IFlyMSC.h"这个头文件，没有智能提示，文档里面我好像也没看到，还有我去年写的时候有备份(下面我会狠狠吐槽讯飞)

---

示例代码

- (BOOL)application:(UIApplication *)application didFinishLaunchingWithOptions:(NSDictionary *)launchOptions {

    //Appid是应用的身份信息，具有唯一性，初始化时必须要传入Appid。
    NSString *initString = [[NSString alloc] initWithFormat:@"appid=%@",xunfeiID];
    [IFlySpeechUtility createUtility:initString];
}

//
//  MMusicViewController.m
//  MusicPlayer
//
//  Created by HZhenF on 2018/2/18.
//  Copyright © 2018年 LiuZhu. All rights reserved.
//

#import "MMusicViewController.h"
#import "IFlyMSC/IFlyMSC.h"
#import "ISRDataHelper.h"

@interface MMusicViewController ()<IFlySpeechRecognizerDelegate>
//不带界面的识别对象
@property (nonatomic, strong) IFlySpeechRecognizer *iFlySpeechRecognizer;
/**生成的字符*/
@property(nonatomic,strong) NSString *resultStringFromJson;
/**当前是否可以进行录音*/
@property(nonatomic,assign) BOOL isStartRecord;
/**是否已经开始播放*/
@property(nonatomic,assign) BOOL ishadStart;

@end

@implementation MMusicViewController

#pragma mark - 懒加载

-(IFlySpeechRecognizer *)iFlySpeechRecognizer
{
    if (!_iFlySpeechRecognizer) {
        //创建语音识别对象
        _iFlySpeechRecognizer = [IFlySpeechRecognizer sharedInstance];
        //设置识别参数
        //设置为听写模式
        [_iFlySpeechRecognizer setParameter:@"iat" forKey:[IFlySpeechConstant IFLY_DOMAIN]];
        //asr_audio_path 是录音文件名，设置 value 为 nil 或者为空取消保存，默认保存目录在 Library/cache 下。
        [_iFlySpeechRecognizer setParameter:@"iat.pcm" forKey:[IFlySpeechConstant ASR_AUDIO_PATH]];
        //设置最长录音时间:60秒
        [_iFlySpeechRecognizer setParameter:@"-1" forKey:[IFlySpeechConstant SPEECH_TIMEOUT]];
        //设置语音后端点:后端点静音检测时间，即用户停止说话多长时间内即认为不再输入， 自动停止录音
        [_iFlySpeechRecognizer setParameter:@"10000" forKey:[IFlySpeechConstant VAD_EOS]];
        //设置语音前端点:静音超时时间，即用户多长时间不说话则当做超时处理
        [_iFlySpeechRecognizer setParameter:@"5000" forKey:[IFlySpeechConstant VAD_BOS]];
        //网络等待时间
        [_iFlySpeechRecognizer setParameter:@"2000" forKey:[IFlySpeechConstant NET_TIMEOUT]];
        //设置采样率，推荐使用16K
        [_iFlySpeechRecognizer setParameter:@"16000" forKey:[IFlySpeechConstant SAMPLE_RATE]];
        //设置语言
        [_iFlySpeechRecognizer setParameter:@"zh_cn" forKey:[IFlySpeechConstant LANGUAGE]];
        //设置方言
        [_iFlySpeechRecognizer setParameter:@"mandarin" forKey:[IFlySpeechConstant ACCENT]];
        //设置是否返回标点符号
        [_iFlySpeechRecognizer setParameter:@"0" forKey:[IFlySpeechConstant ASR_PTT]];
        //设置代理
        _iFlySpeechRecognizer.delegate = self;
    }
    return _iFlySpeechRecognizer;
}


#pragma mark - 系统方法

- (void)viewDidLoad {
    [super viewDidLoad];

    self.view.backgroundColor = [UIColor whiteColor];


    self.isStartRecord = YES;
    //初始化字符串,否则无法拼接
    self.resultStringFromJson = @"";


    CGRect btnRect = CGRectMake((ZFScreenW - 100)*0.5, 200, 100, 100);
    UIButton *btn = [[UIButton alloc] initWithFrame:btnRect];
    btn.backgroundColor = [UIColor orangeColor];
    [btn setTitle:@"语音" forState:UIControlStateNormal];
    [self.view addSubview:btn];

    UILongPressGestureRecognizer *longpress = [[UILongPressGestureRecognizer alloc] initWithTarget:self action:@selector(longPressActionFromLongPressBtn:)];
    longpress.minimumPressDuration = 0.1;
    [btn addGestureRecognizer:longpress];
}


#pragma mark - 自定义方法

-(void)longPressActionFromLongPressBtn:(UILongPressGestureRecognizer *)longPress
{
//    NSLog(@"AIAction");
    CGPoint currentPoint = [longPress locationInView:longPress.view];

    if (self.isStartRecord) {
        self.resultStringFromJson = @"";

        //启动识别服务
        [self.iFlySpeechRecognizer startListening];
        self.isStartRecord = NO;
        self.ishadStart = YES;
        //开始声音动画
//        [self TipsViewShowWithType:@"start"];
    }
    //如果上移的距离大于60,就提示放弃本次录音
    if (currentPoint.y < -60) {
        //变成取消发送图片
//        [self TipsViewShowWithType:@"cancel"];
        self.ishadStart = NO;
    }
    else
    {
        if (self.ishadStart == NO) {
            //开始声音动画
//            [self TipsViewShowWithType:@"start"];
            self.ishadStart = YES;
        }
    }
    if (longPress.state == UIGestureRecognizerStateEnded) {
        self.isStartRecord = YES;
        if (currentPoint.y < -60) {
            [self.iFlySpeechRecognizer cancel];
        }
        else
        {
            [self.iFlySpeechRecognizer stopListening];
        }
        //让声音播放动画消失
//        [self TipsViewShowWithType:@"remove"];
    }

}


//IFlySpeechRecognizerDelegate协议实现
//识别结果返回代理
- (void) onResults:(NSArray *) results isLast:(BOOL)isLast{
    NSMutableString *resultString = [[NSMutableString alloc] init];
    NSDictionary *dic = results[0];
    for (NSString *key in dic) {
        [resultString appendFormat:@"%@",key];
    }

    //持续拼接语音内容
    self.resultStringFromJson = [self.resultStringFromJson stringByAppendingString:[ISRDataHelper stringFromJson:resultString]];
    NSLog(@"self.resultStringFromJson = %@",self.resultStringFromJson);
}
//识别会话结束返回代理
- (void)onError: (IFlySpeechError *) error{
    NSLog(@"error = %@",[error description]);
}

//停止录音回调
-(void)onEndOfSpeech
{
    self.isStartRecord = YES;
}

//开始录音回调
-(void)onBeginOfSpeech
{
    //    NSLog(@"onbeginofspeech");
}

//音量回调函数
-(void)onVolumeChanged:(int)volume
{

}

//会话取消回调
-(void)onCancel
{
    //    NSLog(@"取消本次录音");
}


@end

//
//  ISRDataHelper.h
//  HYAIRobotSDK
//
//  Created by HZhenF on 2017/9/7.
//  Copyright © 2017年 GZHYTechnology. All rights reserved.
//

#import <Foundation/Foundation.h>

@interface ISRDataHelper : NSObject

// 解析命令词返回的结果
+ (NSString*)stringFromAsr:(NSString*)params;

/**
 解析JSON数据
 ****/
+ (NSString *)stringFromJson:(NSString*)params;//


/**
 解析语法识别返回的结果
 ****/
+ (NSString *)stringFromABNFJson:(NSString*)params;

@end

//
//  ISRDataHelper.m
//  HYAIRobotSDK
//
//  Created by HZhenF on 2017/9/7.
//  Copyright © 2017年 GZHYTechnology. All rights reserved.
//

#import "ISRDataHelper.h"

@implementation ISRDataHelper


/**
 解析命令词返回的结果
 ****/
+ (NSString*)stringFromAsr:(NSString*)params;
{
    NSMutableString * resultString = [[NSMutableString alloc]init];
    NSString *inputString = nil;

    NSArray *array = [params componentsSeparatedByString:@"\n"];

    for (int  index = 0; index < array.count; index++)
    {
        NSRange range;
        NSString *line = [array objectAtIndex:index];

        NSRange idRange = [line rangeOfString:@"id="];
        NSRange nameRange = [line rangeOfString:@"name="];
        NSRange confidenceRange = [line rangeOfString:@"confidence="];
        NSRange grammarRange = [line rangeOfString:@" grammar="];

        NSRange inputRange = [line rangeOfString:@"input="];

        if (confidenceRange.length == 0 || grammarRange.length == 0 || inputRange.length == 0 )
        {
            continue;
        }

        //check nomatch
        if (idRange.length!=0) {
            NSUInteger idPosX = idRange.location + idRange.length;
            NSUInteger idLength = nameRange.location - idPosX;
            range = NSMakeRange(idPosX,idLength);
            NSString *idValue = [[line substringWithRange:range]
                                 stringByTrimmingCharactersInSet: [NSCharacterSet whitespaceAndNewlineCharacterSet] ];
            if ([idValue isEqualToString:@"nomatch"]) {
                return @"";
            }
        }

        //Get Confidence Value
        NSUInteger confidencePosX = confidenceRange.location + confidenceRange.length;
        NSUInteger confidenceLength = grammarRange.location - confidencePosX;
        range = NSMakeRange(confidencePosX,confidenceLength);


        NSString *score = [line substringWithRange:range];

        NSUInteger inputStringPosX = inputRange.location + inputRange.length;
        NSUInteger inputStringLength = line.length - inputStringPosX;

        range = NSMakeRange(inputStringPosX , inputStringLength);
        inputString = [line substringWithRange:range];

        [resultString appendFormat:@"%@ 置信度%@\n",inputString, score];
    }

    return resultString;

}

/**
 解析听写json格式的数据
 params例如：
 {"sn":1,"ls":true,"bg":0,"ed":0,"ws":[{"bg":0,"cw":[{"w":"白日","sc":0}]},{"bg":0,"cw":[{"w":"依山","sc":0}]},{"bg":0,"cw":[{"w":"尽","sc":0}]},{"bg":0,"cw":[{"w":"黄河入海流","sc":0}]},{"bg":0,"cw":[{"w":"。","sc":0}]}]}
 ****/
+ (NSString *)stringFromJson:(NSString*)params
{
    if (params == NULL) {
        return nil;
    }

    NSMutableString *tempStr = [[NSMutableString alloc] init];
    NSDictionary *resultDic  = [NSJSONSerialization JSONObjectWithData:    //返回的格式必须为utf8的,否则发生未知错误
                                [params dataUsingEncoding:NSUTF8StringEncoding] options:kNilOptions error:nil];

    if (resultDic!= nil) {
        NSArray *wordArray = [resultDic objectForKey:@"ws"];

        for (int i = 0; i < [wordArray count]; i++) {
            NSDictionary *wsDic = [wordArray objectAtIndex: i];
            NSArray *cwArray = [wsDic objectForKey:@"cw"];

            for (int j = 0; j < [cwArray count]; j++) {
                NSDictionary *wDic = [cwArray objectAtIndex:j];
                NSString *str = [wDic objectForKey:@"w"];
                [tempStr appendString: str];
            }
        }
    }
    return tempStr;
}


/**
 解析语法识别返回的结果
 ****/
+ (NSString *)stringFromABNFJson:(NSString*)params
{
    if (params == NULL) {
        return nil;
    }
    NSMutableString *tempStr = [[NSMutableString alloc] init];
    NSDictionary *resultDic  = [NSJSONSerialization JSONObjectWithData:
                                [params dataUsingEncoding:NSUTF8StringEncoding] options:kNilOptions error:nil];

    NSArray *wordArray = [resultDic objectForKey:@"ws"];
    for (int i = 0; i < [wordArray count]; i++) {
        NSDictionary *wsDic = [wordArray objectAtIndex: i];
        NSArray *cwArray = [wsDic objectForKey:@"cw"];

        for (int j = 0; j < [cwArray count]; j++) {
            NSDictionary *wDic = [cwArray objectAtIndex:j];
            NSString *str = [wDic objectForKey:@"w"];
            NSString *score = [wDic objectForKey:@"sc"];
            [tempStr appendString: str];
            [tempStr appendFormat:@" 置信度:%@",score];
            [tempStr appendString: @"\n"];
        }
    }
    return tempStr;
}

@end

---

小吐槽

在这里，我要狠狠地吐槽一番讯飞语音开放平台，产品功能怎么样我不说，单单是开发文档的说明我就很想骂人了。以前的开发文档还会有详细的说明，现在SDK也更新了，对应的接口也少了，连Demo也很简单了。但是！能不能别告诉我，我要设置其他的东西，好歹也说明现在不对外开放了。还有，现在下载版的docsets文档，我知道你们的本意是方便iOS开发者导入Document目录下，直接方便开发者查看文档，呵呵，我还不如看以前你们给的文档，要啥有啥。早知如此，我还不如直接去用百度的。本来是想抱着尝试的态度，对你们更新的SDK摸索研究一下，现在，我彻底失望了。虽然是完成了任务，成功使用了语音听写功能，老实说，我太失望了。