一、引言

在这篇博客中，我将详细介绍如何在旭日X3派环境中使用讯飞语音SDK进行机器人语音合成。

以下为具体系统架构图，通过构建语音合成节点接受ros话题，将话题内容合成为语音文件。主要难点在于讯飞并不直接提供linux ARM的SDK和相关支持，并且程序的编译需要诸多.so

动态链接库文件和较为复杂的编译流程。讯飞提供了x64和x86环境下的编译脚本64bit_make.sh，Makefile。我们可以将其转换为CMakeLists.txt使源码编译为ros节点。


系统架构图

二、语音合成功能实现

注册并登录讯飞开放平台下载linux 语音合成SDK：讯飞开放平台-以语音交互为核心的人工智能开放平台 (xfyun.cn)

《SDK目录结构一览》

    • bin：
      • msc（生成msc日志）
      • 相关资源文件
    • doc：
      • 相关技术文档
    • include:
      • 调用SDK所需头文件
    • libs：
      • x86/libmsc.so（32位动态库）
      • x64/libmsc.so（64位动态库）
    • samples：
      • tts_offline_sample（离线语音合成示例）

（注意这个文件结构，之后创建ROS节点时仍需相同文件结构）

讯飞语音合成主要API调用流程如下图所示：

img

然后按照图示流程进行C语言编程：

#include <stdlib.h>
#include <stdio.h>
#include <unistd.h>
#include <errno.h>
#include "../../include/qtts.h"
#include "../../include/msp_cmn.h"
#include "../../include/msp_errors.h"
typedef int SR_DWORD;
typedef short int SR_WORD ;

/* 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  
};
/* 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("\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;
}

int main(int argc, char* argv[])
{
	int         ret                  = MSP_SUCCESS;
	const char* login_params         = "appid = bc29f759, 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 = "engine_type = local,voice_name=xiaoyan, text_encoding = UTF8, tts_res_path = fo|res/tts/xiaoyan.jet;fo|res/tts/common.jet, 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                 = "原神启动"; //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
	}

	printf("\n###########################################################################\n");
	printf("## 语音合成开始（Text To Speech，TTS）##\n");
	printf("###########################################################################\n\n");

	/* 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");

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

	return 0;
}

 在libs目录下解压（arm64 .so文件）压缩包——有需要的评论区留下你的邮箱地址

 打开simple目录下的Makefile将其修改为：

#common makefile header

DIR_INC = ../../include
DIR_BIN = ../../bin
DIR_LIB = ../../libs

TARGET	= tts_offline_sample
BIN_TARGET = $(DIR_BIN)/$(TARGET)

CROSS_COMPILE = 
CFLAGS = -g -Wall -I$(DIR_INC)

ifdef LINUX64
LDFLAGS := -L$(DIR_LIB)/arm64
else
LDFLAGS := -L$(DIR_LIB)/arm32 
endif
LDFLAGS += -lmsc -lrt -ldl -lpthread -lstdc++

OBJECTS := $(patsubst %.c,%.o,$(wildcard *.c))

$(BIN_TARGET) : $(OBJECTS)
	$(CROSS_COMPILE)gcc $(CFLAGS) $^ -o $@ $(LDFLAGS)

%.o : %.c
	$(CROSS_COMPILE)gcc -c $(CFLAGS) $< -o $@
clean:
	@rm -f *.o $(BIN_TARGET)

.PHONY:clean

#common makefile foot

修改64bit_make.sh

#编译64位可执行文件
make clean;make LINUX64=1
#设置libmsc.so库搜索路径
export LD_LIBRARY_PATH=$(pwd)/../../libs/arm64/

运行64bit_make.sh，编译生成程序在bin目录下

运行语音合成程序

生成的wav音频文件同样在bin目录下

三、ROS节点建立

这里我们建立一个功能包：包名：yuyin，节点名：yuyin_node

文件结构：

在yuyin/src新建tts_offline_sample.c文件

tts_offline_sample.c代码：

/*
* 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
* 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
* 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。
*/

#include <stdlib.h>
#include <stdio.h>
#include <unistd.h>
#include <errno.h>
#include <cstring>
#include "ros/ros.h"
#include "std_msgs/String.h"
#include "../include/qtts.h"
#include "../include/msp_cmn.h"
#include "../include/msp_errors.h"
typedef int SR_DWORD;
typedef short int SR_WORD ;



/* 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  
};

/* 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("\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;
}

void chatterCallback(const std_msgs::String::ConstPtr& msg)
{
  ROS_INFO("I received: [%s]", msg->data.c_str());
  	int         ret                  = MSP_SUCCESS;
	const char* login_params         = "appid = bc29f759, work_dir = .";//登录参数,appid与msc库绑定,请勿随意改动
	const char* session_begin_params = "engine_type = local,voice_name=xiaoyan, text_encoding = UTF8, tts_res_path = fo|res/tts/xiaoyan.jet;fo|res/tts/common.jet, 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                 = msg->data.c_str(); //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
	}

	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");
}

int main(int argc, char* argv[])
{
	// 初始化ROS节点
  ros::init(argc, argv, "listener");

  // 创建节点句柄
  ros::NodeHandle n;

  // 创建一个Subscriber，订阅名为'tts'的话题，注册回调函数chatterCallback
  ros::Subscriber sub = n.subscribe("tts", 1000, chatterCallback);

    // 循环等待回调函数
  ros::spin();

	return 0;
}

 

修改Cmake

cmake_minimum_required(VERSION 3.0.2)
project(yuyin)


find_package(catkin REQUIRED COMPONENTS
  roscpp
  rospy
  std_msgs
)

catkin_package()


include_directories(
  ../include  # This is equivalent to your DIR_INC
  ${catkin_INCLUDE_DIRS}
)




link_directories(/home/test/src/yuyin/libs/arm64)  # Add the path to your libraries

add_executable(${PROJECT_NAME}_node src/tts_offline_sample.cpp)

target_link_libraries(${PROJECT_NAME}_node
  msc  # These are equivalent to the libraries you specified in LDFLAGS
  rt
  dl
  pthread
  stdc++
  ${catkin_LIBRARIES}
)

install(TARGETS ${PROJECT_NAME}_node
  RUNTIME DESTINATION ${CATKIN_PACKAGE_BIN_DESTINATION}
)

编译测试一下：

大功告成！以后ROS也会大声喊原神，启动！