百度语音SDK开发，姐妹篇~iOS-Demo
iOS语音识别demo-基于百度语音SDK
第一次在CSDN写博客，小小记录一下。哈哈
一、概述
本文主要记录“离在线语音识别”和“语音唤醒”功能。基于百度语音iOS-SDK进行开发。为何不用“讯飞”，额，因为百度免费。
二、SDK下载及环境配置
2.1. 创建应用
a、登录后台创建应用：https://console.bce.baidu.com/ai/#/ai/speech/app/create
b、创建成功后，会生成三个参数：“API_KEY”、“SECRET_KEY”以及“APP_ID”
2.2. 下载SDK
百度语音开放平台有两个入口（估计是后面百度语音开放平台集成到百度AI去了）：
http://yuyin.baidu.com/
https://ai.baidu.com/sdk#asr（推荐）
下载下来的SDK解压后如下：
①SDK解压后的根目录 ②SDK文档 ③头文件 ④百度语音静态库 ⑤语音资源文件 ⑥官方demo
2.3 集成到项目
添加静态库及资源文件
添加依赖库
- libc++.tbd
- libstdc++.6.0.9.tbd
- libz.1.2.5.tbd
- libiconv.2.4.0.tbd
- AudioToolbox
- AVFoundation
- CoreTelephony
- SystemConfiguration
添加头文件
识别相关
#import "BDSEventManager.h"
#import "BDSASRDefines.h"
#import "BDSASRParameters.h"唤醒相关
#import "BDSWakeupDefines.h"
#import "BDSWakeupParameters.h"替换“API_KEY”、“SECRET_KEY”以及“APP_ID”
NSString* const API_KEY = @"your API_KEY";
NSString* const SECRET_KEY = @"your SECRET_KEY";
NSString* const APP_ID = @"your APP_ID";三、简单实用
3.1、语音识别
3.1.1 初始化
- 初始化语音识别对象
self.asrEventManager = [BDSEventManager createEventManagerWithName:BDS_ASR_NAME];- 配置语音识别对象
- (void)initAsrEventManager{
// 设置语音识别代理
[self.asrEventManager setDelegate:self];
// 参数配置：在线身份验证
[self.asrEventManager setParameter:@[_API_KEY,_SECRET_KEY] forKey:BDS_ASR_API_SECRET_KEYS];
//设置 APPID
[self.asrEventManager setParameter:_APP_ID forKey:BDS_ASR_OFFLINE_APP_CODE];
// 参数设置：识别策略为离在线并行
[self.asrEventManager setParameter:@(EVR_STRATEGY_BOTH) forKey:BDS_ASR_STRATEGY];
}- 配置端点检测
/**
检测更加精准，抗噪能力强，响应速度较慢
*/
- (void)configModelVAD {
NSString *modelVAD_filepath = [[NSBundle mainBundle] pathForResource:@"bds_easr_basic_model" ofType:@"dat"];
[self.asrEventManager setParameter:modelVAD_filepath forKey:BDS_ASR_MODEL_VAD_DAT_FILE];
[self.asrEventManager setParameter:@(YES) forKey:BDS_ASR_ENABLE_MODEL_VAD];
}
/**
提供基础检测功能，性能高，响应速度快
*/
- (void)configDNNMFE {
NSString *mfe_dnn_filepath = [[NSBundle mainBundle] pathForResource:@"bds_easr_mfe_dnn" ofType:@"dat"];
[self.asrEventManager setParameter:mfe_dnn_filepath forKey:BDS_ASR_MFE_DNN_DAT_FILE];
NSString *cmvn_dnn_filepath = [[NSBundle mainBundle] pathForResource:@"bds_easr_mfe_cmvn" ofType:@"dat"];
[self.asrEventManager setParameter:cmvn_dnn_filepath forKey:BDS_ASR_MFE_CMVN_DAT_FILE];
}3.1.2 开始识别
[self.asrEventManager sendCommand:BDS_ASR_CMD_START];3.1.3 停止识别
[self.asrEventManager sendCommand:BDS_ASR_CMD_STOP];3.1.4 取消识别
[self.asrEventManager sendCommand:BDS_ASR_CMD_CANCEL];3.1.5 状态及数据回调
#pragma mark - < BDSClientASRDelegate >
- (void)VoiceRecognitionClientWorkStatus:(int)workStatus obj:(id)aObj {
switch (workStatus) {
case EVoiceRecognitionClientWorkStatusStartWorkIng: {
XFLog(@"EVoiceRecognitionClientWorkStatusStartWorkIng:识别工作开始，开始采集及处理数据");
XFLog(@"%@",[self parseLogToDic:aObj]);
if ([self.delegate respondsToSelector:@selector(onAsrReady)]) {
[self.delegate onAsrReady];
}
}
break;
case EVoiceRecognitionClientWorkStatusNewRecordData: {
XFLog(@"EVoiceRecognitionClientWorkStatusNewRecordData:录音数据回调");
XFLog(@"%@",aObj);
if ([self.delegate respondsToSelector:@selector(onWorking)]) {
[self.delegate onWorking];
}
}
break;
case EVoiceRecognitionClientWorkStatusStart: {
XFLog(@"EVoiceRecognitionClientWorkStatusStart:检测到用户开始说话");
XFLog(@"%@",aObj);
if ([self.delegate respondsToSelector:@selector(onAsrBegin)]) {
[self.delegate onAsrBegin];
}
}
break;
case EVoiceRecognitionClientWorkStatusEnd: {
XFLog(@"EVoiceRecognitionClientWorkStatusEnd:本地声音采集结束，等待识别结果返回并结束录音");
XFLog(@"%@",aObj);
if ([self.delegate respondsToSelector:@selector(onAsrEnd)]) {
[self.delegate onAsrEnd];
}
}
break;
case EVoiceRecognitionClientWorkStatusFlushData: {
XFLog(@"EVoiceRecognitionClientWorkStatusFlushData:连续上屏");
XFLog(@"%@",aObj);
if (self.timer.valid) {
[self.timer invalidate];
self.timer = nil;
}
if ([self.delegate respondsToSelector:@selector(onAsrPartialResult:originResult:)]) {
[self.delegate onAsrPartialResult:[self getResultStrForObject:aObj] originResult:aObj];
}
}
break;
case EVoiceRecognitionClientWorkStatusFinish: {
XFLog(@"EVoiceRecognitionClientWorkStatusFinish:语音识别功能完成，服务器返回正确结果");
XFLog(@"%@",[self getDescriptionForDic:aObj]);
if ([self.delegate respondsToSelector:@selector(onAsrFinalResult:originResult:)]) {
[self.delegate onAsrFinalResult:[self getResultStrForObject:aObj] originResult:aObj];
}
}
break;
case EVoiceRecognitionClientWorkStatusMeterLevel: {
XFLog(@"EVoiceRecognitionClientWorkStatusMeterLevel:当前音量回调");
XFLog(@"%@",aObj);
if ([self.delegate respondsToSelector:@selector(onAsrMeterLevel:)]) {
[self.delegate onAsrMeterLevel:[self getVolumeByObject:aObj]];
}
}
break;
case EVoiceRecognitionClientWorkStatusCancel: {
XFLog(@"EVoiceRecognitionClientWorkStatusCancel:用户取消");
XFLog(@"%@",aObj);
if ([self.delegate respondsToSelector:@selector(onAsrCacel)]) {
[self.delegate onAsrCacel];
}
}
break;
case EVoiceRecognitionClientWorkStatusError: {
XFLog(@"EVoiceRecognitionClientWorkStatusError:发生错误");
XFLog(@"%@",aObj);
NSString *errMsg = [self getErrorMsgWithError:aObj];
if ([self.delegate respondsToSelector:@selector(onAsrFinishError:errorMsg:)]) {
[self.delegate onAsrFinishError:(NSError *)aObj errorMsg:errMsg];
}
}
break;
case EVoiceRecognitionClientWorkStatusLoaded: {
XFLog(@"EVoiceRecognitionClientWorkStatusLoaded:离线引擎加载完成");
XFLog(@"%@",aObj);
if ([self.delegate respondsToSelector:@selector(onOfflineLoaded)]) {
[self.delegate onOfflineLoaded];
}
}
break;
case EVoiceRecognitionClientWorkStatusUnLoaded: {
XFLog(@"EVoiceRecognitionClientWorkStatusUnLoaded:离线引擎卸载完成");
XFLog(@"%@",aObj);
if ([self.delegate respondsToSelector:@selector(onOfflineUnLoaded)]) {
[self.delegate onOfflineUnLoaded];
}
}
break;
case EVoiceRecognitionClientWorkStatusChunkThirdData: {
XFLog(@"EVoiceRecognitionClientWorkStatusChunkThirdData:识别结果中的第三方数据");
XFLog(@"%@",aObj);
}
break;
case EVoiceRecognitionClientWorkStatusChunkNlu: {
XFLog(@"EVoiceRecognitionClientWorkStatusChunkNlu:识别结果中的语义结果");
XFLog(@"%@",aObj);
}
break;
case EVoiceRecognitionClientWorkStatusChunkEnd: {
XFLog(@"EVoiceRecognitionClientWorkStatusChunkEnd:识别过程结束");
XFLog(@"%@",aObj);
}
break;
case EVoiceRecognitionClientWorkStatusFeedback: {
XFLog(@"EVoiceRecognitionClientWorkStatusFeedback:识别过程反馈的打点数据");
XFLog(@"%@",aObj);
}
break;
case EVoiceRecognitionClientWorkStatusRecorderEnd: {
XFLog(@"EVoiceRecognitionClientWorkStatusRecorderEnd:录音机关闭，页面跳转需检测此时间，规避状态条 (iOS)");
XFLog(@"%@",aObj);
}
break;
case EVoiceRecognitionClientWorkStatusLongSpeechEnd: {
XFLog(@"EVoiceRecognitionClientWorkStatusLongSpeechEnd:长语音结束状态");
XFLog(@"%@",aObj);
}
break;
default:
break;
}
}3.1.6 处理接口回调数据
#pragma mark 3.2、语音唤醒
3.2.1 初始化唤醒对象
self.wakeupEventManager = [BDSEventManager createEventManagerWithName:BDS_WAKEUP_NAME];- (void)configWakeUpManager{
// 设置语音唤醒代理
[self.wakeupEventManager setDelegate:self];
// 参数配置：离线授权APPID
[self.wakeupEventManager setParameter:_APP_ID forKey:BDS_WAKEUP_APP_CODE];
// 参数配置：唤醒语言模型文件路径, 默认文件名为 bds_easr_basic_model.dat
NSString* dat = [[NSBundle mainBundle] pathForResource:@"bds_easr_basic_model" ofType:@"dat"];
[self.wakeupEventManager setParameter:dat forKey:BDS_WAKEUP_DAT_FILE_PATH];
//设置唤醒词文件路径
// 默认的唤醒词文件为"bds_easr_wakeup_words.dat"，包含的唤醒词为"百度一下"
// 如需自定义唤醒词，请在 http://ai.baidu.com/tech/speech/wake 中评估并下载唤醒词文件，替换此参数
NSString* words = [[NSBundle mainBundle] pathForResource:@"xiaomingwakeup" ofType:@"dms"];
[self.wakeupEventManager setParameter:words forKey:BDS_WAKEUP_WORDS_FILE_PATH];
}3.2.2 开始唤醒
- (void)startWakeup{
[self.wakeupEventManager setParameter:nil forKey:BDS_WAKEUP_AUDIO_FILE_PATH];
[self.wakeupEventManager setParameter:nil forKey:BDS_WAKEUP_AUDIO_INPUT_STREAM];
// 发送指令：加载语音唤醒引擎
[self.wakeupEventManager sendCommand:BDS_WP_CMD_LOAD_ENGINE];
// 发送指令：启动唤醒
[self.wakeupEventManager sendCommand:BDS_WP_CMD_START];
}3.2.3 关闭唤醒
- (void)stopWakeup{
[self.wakeupEventManager sendCommand:BDS_WP_CMD_STOP];
}3.2.4 唤醒接口回调
- (void)WakeupClientWorkStatus:(int)workStatus obj:(id)aObj{
switch (workStatus) {
case EWakeupEngineWorkStatusStarted: {
XFLog(@"EWakeupEngineWorkStatusStarted");
if ([self.delegate respondsToSelector:@selector(wakeEngineStart)]) {
[self.delegate wakeEngineStart];
}
}
break;
case EWakeupEngineWorkStatusStopped: {
XFLog(@"EWakeupEngineWorkStatusStopped");
if ([self.delegate respondsToSelector:@selector(wakeEngineStop)]) {
[self.delegate wakeEngineStop];
}
}
break;
case EWakeupEngineWorkStatusLoaded: {
XFLog(@"EWakeupEngineWorkStatusLoaded");
if ([self.delegate respondsToSelector:@selector(wakeEngineLoaded)]) {
[self.delegate wakeEngineLoaded];
}
}
break;
case EWakeupEngineWorkStatusUnLoaded: {
XFLog(@"EWakeupEngineWorkStatusUnLoaded");
if ([self.delegate respondsToSelector:@selector(wakeEngineUnLoaded)]) {
[self.delegate wakeEngineUnLoaded];
}
}
break;
case EWakeupEngineWorkStatusTriggered: { // 命中唤醒词
XFLog(@"EWakeupEngineWorkStatusTriggered:%@",(NSString *)aObj);
if ([self.delegate respondsToSelector:@selector(wakeEngineTriggered:)]) {
[self.delegate wakeEngineTriggered:(NSString *)aObj];
}
}
break;
case EWakeupEngineWorkStatusError: {
XFLog(@"EWakeupEngineWorkStatusError : %@",(NSString *)aObj);
if ([self.delegate respondsToSelector:@selector(wakeEngineError:)]) {
[self.delegate wakeEngineError:(NSError *)aObj];
}
}
break;
default:
break;
}
}3.3、如何自定义唤醒词及离线识别语法
3.3.1 自定义唤醒词
详见3.2.1初始化唤醒对象的“configWakeUpManager（）”方法
3.3.2 自定义离线识别语法
自定义离线语法并下载资源文件
http://speech.baidu.com/asr将自定义的离线语法添加进项目
加载并启动“离线识别引擎”
- (void)loadOfflineRecogEngine{
// 参数设置：离线识别引擎类型
[self.asrEventManager setParameter:@(EVR_OFFLINE_ENGINE_GRAMMER) forKey:BDS_ASR_OFFLINE_ENGINE_TYPE];
// 参数配置：命令词引擎语法文件路径。 请在 (官网)[http://speech.baidu.com/asr] 参考模板定义语法，下载语法文件后，替换BDS_ASR_OFFLINE_ENGINE_GRAMMER_FILE_PATH参数
NSString *offlineGrammerPath = [[NSBundle mainBundle] pathForResource:@"xfdemo_speech_grammar" ofType:@"dat"];
[self.asrEventManager setParameter:offlineGrammerPath forKey:BDS_ASR_OFFLINE_ENGINE_GRAMMER_FILE_PATH];
// // 参数配置：命令词引擎语言模型文件路径，用于动态更新离线语法
NSString *offlineDatPath = [[NSBundle mainBundle] pathForResource:@"bds_easr_basic_model" ofType:@"dat"];
[self.asrEventManager setParameter:offlineDatPath forKey:BDS_ASR_OFFLINE_ENGINE_DAT_FILE_PATH];
// 发送指令：加载离线引擎
[self.asrEventManager sendCommand:BDS_ASR_CMD_LOAD_ENGINE];
}第一次写博客，欢迎大家批评指点。