【UE5/UE4】超详细教程接入科大讯飞语音唤醒SDK并初始持久监听（10102错误码解决）

【ue5/ue4】超详细教程接入科大讯飞语音唤醒sdk并初始持久监听（10102错误码解决）

先导

windows环境下
**ue版本为ue4.27及以下
你需要具备一定的c++基础，或全程一对一对照我的代码编写你的代码
使用offline voice recognition插件作为录音插件（仅做录音插件使用）
基于https://github.com/zhangmei126/xunfei二次开发
语音识别部分参考csdnue4如何接入科大讯飞的语音识别
在此基础上增加了语音唤醒功能，实际上语音唤醒与上述文章中是两个模块
由于讯飞插件的使用需要调用msplogin，也就是需要先注册

其插件中的speechinit（）方法已经为我们注册好了，如果可以自己写注册的话，后述本文语音唤醒部分是不分引擎版本的

语音唤醒环境配置

参考ue4如何接入科大讯飞的语音识别接入科大讯飞sdk以及使用offline voice recognition插件了后，在plugins中确保他们都是开启的状态


你要确保你的sdk下载的是windows版本以及一下sdk文件包

你要确保你的sdk下载后正确导入了且appid已经拥有了正确的配置

你要确保你sdk下载后将windows_awaken_exp1227_iat1226_af12f15d\bin\msc\res\ivw\wakeupresource.jet放置到了你的c盘根目录下

to
如果你是为了10102错误码来此文章：
此处是讯飞的坑，讯飞的wakeupresource.jet的路径必须是绝对路径
c++中也就必须要使用转义符"\\"即
const char* params = "ivw_threshold=0:5, ivw_res_path =fo|c:\\wakeupresource.jet";
至此10102解决了，就这么简单，但是很坑
打开visualstudio

1. 打开讯飞插件文件夹
   
2. 在speechactor.cpp中引入qivw.h
   
   qivw.h语音唤醒具备以下方法
   
3. 在speechactor.h中

#pragma once

#include "speechtask.h"

#include "coreminimal.h"
#include "gameframework/actor.h"
#include "speechactor.generated.h"

uclass()
class xunfei_api aspeechactor : public aactor
{
	generated_body()
private:
	fstring result;
	declare_dynamic_multicast_delegate(fwakeupbufferdelegate);
public:
	// sets default values for this actor's properties
	aspeechactor();

protected:
	// called when the game starts or when spawned
	virtual void beginplay() override;

public:
	
	// called every frame
	virtual void tick(float deltatime) override;

	ufunction(blueprintcallable, category = "xunfei", meta = (displayname = "speechinit", keywords = "speech recognition initialization"))
	void speechinit();

	ufunction(blueprintcallable, category = "xunfei", meta = (displayname = "speechopen", keywords = "speech recognition open"))
	void speechopen();

	ufunction(blueprintcallable, category = "xunfei", meta = (displayname = "speechstop", keywords = "speech recognition stop"))
	void speechstop();

	ufunction(blueprintcallable, category = "xunfei", meta = (displayname = "speechquit", keywords = "speech recognition quit"))
	void speechquit();
	
	ufunction(blueprintcallable, category = "xunfei", meta = (displayname = "speechresult", keywords = "speech recognition getresult"))
	fstring speechresult();
	ufunction(blueprintcallable, category = "xunfei", meta = (displayname = "wakeupstart", keywords = "speech recognition getresult"))
	fstring wakeupstart();
	ufunction(blueprintcallable, category = "xunfei", meta = (displayname = "wakeupend", keywords = "speech recognition getresult"))
	bool wakeupend(fstring sessionid);
	ufunction(blueprintcallable, category = "xunfei", meta = (displayname = "wakeupbuffer", keywords = "speech recognition getresult"))
	bool wakeupbuffer(tarray<uint8> myarray, fstring sessionid);

	//这是一个回调函数
	uproperty(blueprintassignable)
	 fwakeupbufferdelegate onwakeupbuffer;
	
};

4. 在speechactor.cpp中

// fill out your copyright notice in the description page of project settings.
#pragma once

#include "speechactor.h"
#include "xunfei.h"
#include "dom/jsonobject.h"
#include "serialization/jsonreader.h"
#include "serialization/jsonserializer.h"
#include <qivw.h>

// sets default values
aspeechactor::aspeechactor() :
	result{}
{
 	// set this actor to call tick() every frame.  you can turn this off to improve performance if you don't need it.
	primaryactortick.bcanevertick = false;
}



// called when the game starts or when spawned
void aspeechactor::beginplay()
{
	super::beginplay();
}

// called every frame
void aspeechactor::tick(float deltatime)
{
	super::tick(deltatime);

}

void aspeechactor::speechinit()
{
	fautodeleteasynctask<fspeechtask>* speechtask = new fautodeleteasynctask<fspeechtask>();

	if (speechtask)
	{
		speechtask->startbackgroundtask();
	}
	else
	{
		ue_log(xunfeilog, error, text("xunfei task object could not be create !"));
		return;
	}

	ue_log(xunfeilog, log, text("xunfei task stopped !"));
	return;
}

void aspeechactor::speechopen()
{
	xunfeispeech->setstart();
	return;
}

void aspeechactor::speechstop()
{
	xunfeispeech->setstop();
	return;
}

void aspeechactor::speechquit()
{
	xunfeispeech->setquit();
	sleep(300);
	return;
}

fstring aspeechactor::speechresult()
{	
	result = fstring(utf8_to_tchar(xunfeispeech->get_result()));

	fstring lajistring("{\"sn\":2,\"ls\":true,\"bg\":0,\"ed\":0,\"ws\":[{\"bg\":0,\"cw\":[{\"sc\":0.00,\"w\":\"\"}]}]}");

	int32 lajiindex = result.find(*lajistring);

	if (lajiindex != -1)
	{
		result.removefromend(lajistring);
	}

	tsharedptr<fjsonobject> jsonobject;

	tsharedref< tjsonreader<tchar> > reader = tjsonreaderfactory<tchar>::create(result);

	if (fjsonserializer::deserialize(reader, jsonobject))
	{
		result.reset();
		tarray< tsharedptr<fjsonvalue> > temparray = jsonobject->getarrayfield("ws");
		for (auto rs : temparray)
		{
			result.append((rs->asobject()->getarrayfield("cw"))[0]->asobject()->getstringfield("w"));
		}
	}

	ue_log(xunfeilog, log, text("%s"), *result);
	
	return result;
}

// 在 cb_ivw_msg_proc 静态函数中进行事件触发
int cb_ivw_msg_proc(const char* sessionid1, int msg, int param1, int param2, const void* info, void* userdata)
{
	if (msp_ivw_msg_error == msg) //唤醒出错消息
	{
		ue_log(logtemp, warning, text("不在"));
		return 0;
	}
	else if (msp_ivw_msg_wakeup == msg) //唤醒成功消息
	{
		ue_log(logtemp, warning, text("imhere"));
		if (userdata) {
			aspeechactor* mythis = reinterpret_cast<aspeechactor*>(userdata);
			if (mythis) {
				ue_log(logtemp, warning, text("diaoyongle"));
				mythis->onwakeupbuffer.broadcast();
				return 1;
			}
		}
	}
	return 0;
}
fstring aspeechactor::wakeupstart()
{
	int err_code = msp_success;
	const char* params = "ivw_threshold=0:5, ivw_res_path =fo|c:\\wakeupresource.jet";
	int ret = 0;
	const char* sessionid = qivwsessionbegin(null, params, &ret);
	err_code = qivwregisternotify(sessionid,cb_ivw_msg_proc, this);
	if (err_code != msp_success)
	{
		ue_log(logtemp, warning, text("qivwregisternotify failed, error code is: %d"), ret);
	}
	else {
		ue_log(logtemp, warning, text("qivwregisternotify success, error code is: %d"), ret);
	}
	
	if (msp_success != ret)
	{
		
		
		ue_log(logtemp, warning, text("qivwsessionbegin failed, error code is: %d"),ret);
	}
	return fstring(sessionid);
	ue_log(logtemp, warning, text("qivwsessionbegin is working"));
	
	
}

bool aspeechactor::wakeupend(fstring sessionid)
{
	
		int ret = qivwsessionend(tchar_to_ansi(*sessionid) , "normal end");
	if (msp_success != ret)
	{
		ue_log(logtemp, warning, text("qivwsessionend failed, error code is: %d"), ret);
		return false;
	}
	ue_log(logtemp, warning, text("qivwsessionisend"));
	return true;
}


bool aspeechactor::wakeupbuffer(tarray<uint8> bitarray, fstring sessionid)
{
	int ret = 0;

	if (bitarray.num() == 0)
	{
		
		return false;
	}
	else
	{
		int audio_len = bitarray.num();
		int audio_status =2;  // 设置音频状态，这里假设为msp_audio_sample_last

		ret = qivwaudiowrite(tchar_to_ansi(*sessionid), bitarray.getdata(), audio_len, audio_status);
		if (msp_success != ret)
		{
				printf("qivwaudiowrite failed, error code is: %d", ret);
				return false;
		}
		return true;
	}
}

至此准备工作完成
打开我们的蓝图

1. 在确保你的offline voice recognition插件打开的前提下添加vosk插件
   

2. 考虑本文的应用环境需要他初始化时就持续监听
   故而让其开始运行时就开启录音（语音唤醒需要录音文件才能监听）

1. 接下里我们需要注册讯飞，并且在运行结束释放讯飞注册
   
   4.讯飞注册后等待两秒注册我们的wakeup语音唤醒
   
2. 绑定一个唤醒事件

3. 在设置一个1.1秒的定时器

4. 定时器内部（正在语音识别默认值为false）



5. 语音唤醒的自定义事件（唤醒五秒钟后恢复继续监听）

蓝图文件以绑定
author：dacimal
定制化开发联系：19815779273（不闲聊）