ROS学习笔记-ROS语音识别与语音输出[1]

说明：代码部分是基于古月居前辈的例程，在此对胡老师表示感谢！！

语音识别功能框图:

调用过程：

roscore 
rosrun robot_voice iat_publish
rostopic pub /voiceWakeup std_msgs/String "data: '123'"

代码描述:

// 声明Publisher和Subscriber
// 订阅唤醒语音识别的信号
ros::Subscriber wakeUpSub = n.subscribe("voiceWakeup", 1000, WakeUp);   
// 订阅唤醒语音识别的信号    
ros::Publisher voiceWordsPub = n.advertise<std_msgs::String>("voiceWords", 1000); 

订阅话题后执行的回调函数:

void WakeUp(const std_msgs::String::ConstPtr& msg)
{
    printf("waking up\r\n");
    usleep(700*1000);
    wakeupFlag=1;
} 

main()函数框架

int main(int argc, char* argv[])
{
    // 初始化ROS
    ros::init(argc, argv, "voiceRecognition");
    ros::NodeHandle n;
    ros::Rate loop_rate(10);

    // 声明Publisher和Subscriber
    // 订阅唤醒语音识别的信号
    ros::Subscriber wakeUpSub = n.subscribe("voiceWakeup", 1000, WakeUp);   
    // 订阅唤醒语音识别的信号    
    ros::Publisher voiceWordsPub = n.advertise<std_msgs::String>("voiceWords", 1000);  

    ROS_INFO("Sleeping...");
    int count=0;

    int ret = MSP_SUCCESS;
    /* login params, please do keep the appid correct */
    const char* login_params = "appid = 594a7b46, work_dir = .";
    int aud_src = 0; /* from mic or file */

    /*
    * See "iFlytek MSC Reference Manual"
    */
    const char* session_begin_params =
        "sub = iat, domain = iat, language = zh_cn, "
        "accent = mandarin, sample_rate = 16000, "
        "result_type = plain, result_encoding = utf8";

    /* Login first. the 1st arg is username, the 2nd arg is password
     * just set them as NULL. the 3rd arg is login paramertes 
     * */
    ret = MSPLogin(NULL, NULL, login_params);
    if (MSP_SUCCESS != ret)	{
        printf("MSPLogin failed , Error code %d.\n",ret);
        goto exit; // login fail, exit the program
    }

    while(ros::ok())
    {
        // 语音识别唤醒     
        if(wakeupFlag)
        {
            ROS_INFO("Wakeup...");

	        printf("Demo recognizing the speech from microphone\n");
	        printf("Speak in 8 seconds\n");

	        demo_mic(session_begin_params);

	        printf("8 sec passed\n");
            
            wakeupFlag=0;
        }

        // 语音识别完成
        if(resultFlag){
            resultFlag=0;
            std_msgs::String msg;
            msg.data = g_result;
            voiceWordsPub.publish(msg);
        }

        ros::spinOnce();
        loop_rate.sleep();
        count++;
    }
exit:
	MSPLogout(); // Logout...

	return 0;
}

demo_mic()函数

/* demo recognize the audio from microphone */
static void demo_mic(const char* session_begin_params)
{
	int errcode;
	int i = 0;

	struct speech_rec iat;

	struct speech_rec_notifier recnotifier = {
		on_result,
		on_speech_begin,
		on_speech_end
	};

	errcode = sr_init(&iat, session_begin_params, SR_MIC, &recnotifier);
	if (errcode) {
		printf("speech recognizer init failed\n");
		return;
	}
	errcode = sr_start_listening(&iat);
	if (errcode) {
		printf("start listen failed %d\n", errcode);
	}
	/* demo 8 seconds recording */
	while(i++ < 8)
		sleep(1);	//停顿一秒钟
	errcode = sr_stop_listening(&iat);
	if (errcode) {
		printf("stop listening failed %d\n", errcode);
	}
	sr_uninit(&iat);
}

关于 sleep() 函数，可以参考这篇：
https://blog.csdn.net/cxs5534/article/details/105759883
这里使用了自定义的Sleep()
c static void Sleep(size_t ms) { usleep(ms*1000); }

语音输出：

源码阅读：

int main(int argc, char* argv[])
{
	int         ret                  = MSP_SUCCESS;
	const char* login_params         = "appid = 594a7b46, work_dir = .";//登录参数,appid与msc库绑定,请勿随意改动
	/*
	* rdn:           合成音频数字发音方式
	* volume:        合成音频的音量
	* pitch:         合成音频的音调
	* speed:         合成音频对应的语速
	* voice_name:    合成发音人
	* sample_rate:   合成音频采样率
	* text_encoding: 合成文本编码格式
	*
	*/
	//const char* session_begin_params = "voice_name = xiaoyan, text_encoding = utf8, sample_rate = 16000, speed = 50, volume = 50, pitch = 50, rdn = 2";
	//const char* filename             = "tts_sample.wav"; //合成的语音文件名称
	//const char* text                 = "亲爱的用户，您好，这是一个语音合成示例，感谢您对科大讯飞语音技术的支持！科大讯飞是亚太地区最大的语音上市公司，股票代码：002230"; //合成文本

	/* 用户登录 */
	ret = MSPLogin(NULL, NULL, login_params);//第一个参数是用户名，第二个参数是密码，第三个参数是登录参数，用户名和密码可在http://www.xfyun.cn注册获取
	if (MSP_SUCCESS != ret)
	{
		printf("MSPLogin failed, error code: %d.\n", ret);
		//goto exit ;//登录失败，退出登录
        toExit();
	}
	printf("\n###########################################################################\n");
	printf("## 语音合成（Text To Speech，TTS）技术能够自动将任意文字实时转换为连续的 ##\n");
	printf("## 自然语音，是一种能够在任何时间、任何地点，向任何人提供语音信息服务的  ##\n");
	printf("## 高效便捷手段，非常符合信息时代海量数据、动态更新和个性化查询的需求。  ##\n");
	printf("###########################################################################\n\n");

    ros::init(argc,argv,"TextToSpeech");
    ros::NodeHandle n;
    ros::Subscriber sub =n.subscribe("voiceWords", 1000, voiceWordsCallback);
    ros::spin();

exit:
	printf("按任意键退出 ...\n");
	getchar();
	MSPLogout(); //退出登录

	return 0;
}

voiceWordsCallback()函数：

void voiceWordsCallback(const std_msgs::String::ConstPtr& msg)
{
    char cmd[2000];
    const char* text;
    int         ret                  = MSP_SUCCESS;
    const char* session_begin_params = "voice_name = xiaoyan, text_encoding = utf8, sample_rate = 16000, speed = 50, volume = 50, pitch = 50, rdn = 2";
    const char* filename             = "tts_sample.wav"; //合成的语音文件名称

    std::cout<<"I heard :"<<msg->data.c_str()<<std::endl;
    text = msg->data.c_str(); 

    /* 文本合成 */
    printf("开始合成 ...\n");
    ret = text_to_speech(text, filename, session_begin_params);
    if (MSP_SUCCESS != ret)
    {
        printf("text_to_speech failed, error code: %d.\n", ret);
    }
    printf("合成完毕\n");
 	//在终端调用输出语音
    popen("play tts_sample.wav","r");
    sleep(1);
}

文本合成函数　text_to_speech()

　/*　 文本合成　 */
int text_to_speech(const char* src_text, const char* des_path, const char* params)
{
	int          ret          = -1;
	FILE*        fp           = NULL;
	const char*  sessionID    = NULL;
	unsigned int audio_len    = 0;
	wave_pcm_hdr wav_hdr      = default_wav_hdr;
	int          synth_status = MSP_TTS_FLAG_STILL_HAVE_DATA;
	if (NULL == src_text || NULL == des_path)
	{
		printf("params is error!\n");
		return ret;
	}
	fp = fopen(des_path, "wb");
	if (NULL == fp)
	{
		printf("open %s error.\n", des_path);
		return ret;
	}
	/* 开始合成 */
	sessionID = QTTSSessionBegin(params, &ret);
	if (MSP_SUCCESS != ret)
	{
		printf("QTTSSessionBegin failed, error code: %d.\n", ret);
		fclose(fp);
		return ret;
	}
	ret = QTTSTextPut(sessionID, src_text, (unsigned int)strlen(src_text), NULL);
	if (MSP_SUCCESS != ret)
	{
		printf("QTTSTextPut failed, error code: %d.\n",ret);
		QTTSSessionEnd(sessionID, "TextPutError");
		fclose(fp);
		return ret;
	}
	printf("正在合成 ...\n");
	fwrite(&wav_hdr, sizeof(wav_hdr) ,1, fp); //添加wav音频头，使用采样率为16000
	while (1) 
	{
		/* 获取合成音频 */
		const void* data = QTTSAudioGet(sessionID, &audio_len, &synth_status, &ret);
		if (MSP_SUCCESS != ret)
			break;
		if (NULL != data)
		{
			fwrite(data, audio_len, 1, fp);
		    wav_hdr.data_size += audio_len; //计算data_size大小
		}
		if (MSP_TTS_FLAG_DATA_END == synth_status)
			break;
		printf(">");
		usleep(150*1000); //防止频繁占用CPU
	}
	printf("\n");
	if (MSP_SUCCESS != ret)
	{
		printf("QTTSAudioGet failed, error code: %d.\n",ret);
		QTTSSessionEnd(sessionID, "AudioGetError");
		fclose(fp);
		return ret;
	}
	/* 修正wav文件头数据的大小 */
	wav_hdr.size_8 += wav_hdr.data_size + (sizeof(wav_hdr) - 8);
	
	/* 将修正过的数据写回文件头部,音频文件为wav格式 */
	fseek(fp, 4, 0);
	fwrite(&wav_hdr.size_8,sizeof(wav_hdr.size_8), 1, fp); //写入size_8的值
	fseek(fp, 40, 0); //将文件指针偏移到存储data_size值的位置
	fwrite(&wav_hdr.data_size,sizeof(wav_hdr.data_size), 1, fp); //写入data_size的值
	fclose(fp);
	fp = NULL;
	/* 合成完毕 */
	ret = QTTSSessionEnd(sessionID, "Normal");
	if (MSP_SUCCESS != ret)
	{
		printf("QTTSSessionEnd failed, error code: %d.\n",ret);
	}

	return ret;
}

数据结构：

/* wav音频头部格式 */
typedef struct _wave_pcm_hdr
{
	char            riff[4];                // = "RIFF"
	int		size_8;                 // = FileSize - 8
	char            wave[4];                // = "WAVE"
	char            fmt[4];                 // = "fmt "
	int		fmt_size;		// = 下一个结构体的大小 : 16

	short int       format_tag;             // = PCM : 1
	short int       channels;               // = 通道数 : 1
	int		samples_per_sec;        // = 采样率 : 8000 | 6000 | 11025 | 16000
	int		avg_bytes_per_sec;      // = 每秒字节数 : samples_per_sec * bits_per_sample / 8
	short int       block_align;            // = 每采样点字节数 : wBitsPerSample / 8
	short int       bits_per_sample;        // = 量化比特数: 8 | 16

	char            data[4];                // = "data";
	int		data_size;              // = 纯数据长度 : FileSize - 44 
} wave_pcm_hdr;

/* 默认wav音频头部数据 */
wave_pcm_hdr default_wav_hdr = 
{
	{ 'R', 'I', 'F', 'F' },
	0,
	{'W', 'A', 'V', 'E'},
	{'f', 'm', 't', ' '},
	16,
	1,
	1,
	16000,
	32000,
	2,
	16,
	{'d', 'a', 't', 'a'},
	0  
};