前言

语音识别相关算法一般在MATLAB上进行仿真验证与实验，在工程上一般还是在VS中进行实现落地，本系列将介绍语音信号处理在C语言中的一系列应用，后期将以此为基础，再落地移植到嵌入式平台。

今天介绍WAV格式语音存为PCM格式语音的工程应用。两种格式语音的唯一区别在于WAV语音包含文件头信息，而PCM格式语音仅包含语音数据流。语音格式转换也算是一种比较常见的应用，可以让我们更好的理解语音文件的格式信息。

一. 工程实现

打开VS2015
在这里插入图片描述
点击新建一个项目（工程），输入项目名：Wav2pcm

点击“确定”，创建项目

左侧解决方案资源管理器视图中，源文件栏右键添加新建项，

点击C++文件，新建一个源文件Wav2pcm.cpp，点击“添加”

输入代码如下：

// Wav2pcm.cpp
// wav格式语音文件转存为pcm格式语音
// date：2022-5-9 22:32:21
// author : C.S

#include <stdio.h>  
#include <string.h>  
#include <stdlib.h>

#pragma warning (disable :4996)

#define  uint32_t       unsigned int  
#define  uint16_t       unsigned short   


typedef struct _riff_t {
	char riff[4];	/* "RIFF" (ASCII characters) */
	uint32_t len;	/* Length of package (binary, little endian) */
	char wave[4];	/* "WAVE" (ASCII characters) */
} riff_t;

/* The FORMAT chunk */

typedef struct _format_t {
	char  fmt[4];		/* "fmt_" (ASCII characters) */
	uint32_t   len;	/* length of FORMAT chunk (always 0x10) */
	uint16_t  type;		/* codec type*/
	uint16_t channel;	/* Channel numbers (0x01 = mono, 0x02 = stereo) */
	uint32_t   rate;	/* Sample rate (binary, in Hz) */
	uint32_t   bps;	/* Average Bytes Per Second */
	uint16_t blockalign;	/*number of bytes per sample */
	uint16_t bitpspl;	/* bits per sample */
} format_t;

typedef struct _data_t {
	char data[4];   /* "data" (ASCII characters) */
	uint32_t len;  /* length of data */
}data_t;

typedef struct _wav_head
{
	riff_t riff;
	format_t format;
	data_t   data;
}wav_head;




int main(int argc, char **argv)
{

	char *src_file = "test.wav";

	uint32_t pcm_data_size = 0;
	int pcm_channel = 0;
	int pcm_samplerate = 0;

	//以下是为了建立.wav头而准备的变量    
	wav_head header;
	FILE   *fp = NULL, *fpCpy = NULL;

	int flen;



	if ((fp = fopen(src_file, "rb")) == NULL) //读取文件    
	{
		printf("open wav file %s error\n", argv[1]);
		return -1;
	}


	if ((flen = fread(&header.riff.riff, 1, 4, fp)) != 4)
		goto error;

	if (memcmp(&header.riff.riff, "RIFF", 4))
		goto error;

	if ((flen = fread(&header.riff.len, 1, 4, fp)) != 4)
		goto error;

	if ((flen = fread(&header.riff.wave, 1, 4, fp)) != 4)
		goto error;

	if (memcmp(&header.riff.wave, "WAVE", 4))
		goto error;

	if ((flen = fread(&header.format.fmt, 1, 4, fp)) != 4)
		goto error;

	if (memcmp(&header.format.fmt, "fmt ", 4))
		goto error;

	if ((flen = fread(&header.format.len, 1, 4, fp)) != 4)
	{
		goto error;
	}
	if (header.format.len != 0x10)
	{
		printf("header.format.len!=0x10 :value:%d\n", header.format.len);
	}
	if ((flen = fread(&header.format.type, 1, 2, fp)) != 2)
	{
		goto error;
	}
	if (header.format.type != 0x1)
	{
		printf("header.format.type != 0x1 :value:%d\n", header.format.type);
	}
	if ((flen = fread(&header.format.channel, 1, 2, fp)) != 2)
	{
		goto error;
	}
	pcm_channel = header.format.channel;

	if ((flen = fread(&header.format.rate, 1, 4, fp)) != 4)
	{
		goto error;
	}
	pcm_samplerate = header.format.rate;

	if ((flen = fread(&header.format.bps, 1, 4, fp)) != 4)
	{
		goto error;
	}

	if ((flen = fread(&header.format.blockalign, 1, 2, fp)) != 2)
	{
		goto error;
	}

	if ((flen = fread(&header.format.bitpspl, 1, 2, fp)) != 2)
	{
		goto error;
	}

	int p = 0;
	//后面的结构可能不是直接为“data”
	do {
		if ((flen = fread(&header.data.data, 1, 4, fp)) != 4)
		{
			goto error;
		}
		if (memcmp(&header.data.data, "data", 4) == 0)
		{
			printf("find data chunk.");
			if ((flen = fread(&header.data.len, 1, 4, fp)) != 4)
			{
				goto error;
			}
			pcm_data_size = header.data.len;
			break;
		}
		else
			p += 4;

	} while (!feof(fp));


	printf("解析成功，二段PCM数据长度分别为:%d, %d\n", pcm_data_size, header.riff.len - 36);

	int otherdataSize = header.riff.len - 36 - pcm_data_size;
	if (otherdataSize>0)
	{
		printf("多余的数据部分大小：%d , %d", p, otherdataSize);

		char *otherdata = (char *)malloc(otherdataSize + 1);
		memset(otherdata, 0, otherdataSize + 1);
		fseek(fp, -otherdataSize - 8, SEEK_CUR);
		fread(otherdata, 1, otherdataSize, fp);
		*(otherdata + otherdataSize) = '\0';

		fseek(fp, 8, SEEK_CUR);

		printf("   other data:%s", otherdata);
	}
	fseek(fp, otherdataSize + 44, SEEK_SET);

	//fseek(fp, 0, SEEK_CUR);

	unsigned char *buffer = (unsigned char*)malloc((pcm_data_size) * sizeof(char));

	memset(buffer, 0, pcm_data_size);


	int count1 = fread(buffer, 1, pcm_data_size, fp);


	char output[1024] = { 0 };
	sprintf(output, "%d_%d_.pcm", pcm_samplerate, pcm_channel);

	if ((fpCpy = fopen(output, "wb+")))
	{
		//fwrite(buffer, 1, sizeof(buffer), fpCpy);  写了段常犯错的代码
		fwrite(buffer, 1, pcm_data_size, fpCpy);
	}
	else
		goto error;


	free(buffer);
	fclose(fp); //关闭文件    
	fclose(fpCpy);   //关闭文件    

	return 0;

error:
	if (fp)
	{
		fclose(fp);
	}
	if (fpCpy)
	{
		fclose(fpCpy);
	}
	return -1;
}