百度实时语音识别示例教程-基于libwebsockets库，linux c++版本

百度实时语音识别示例教程-基于libwebsockets库，linux c++版本

准备

1. 开通百度关于实时语音项目的内容。成功后，会有以下结果。用到的只有appid和api key。

2. 在百度控制台那边领取免费额度，下图右下角。成功后，会有下图底部的总量10小时余额。

3. 准备要识别的pcm语音文件，可以到百度云sdk页面下载sdk，然后sdk压缩包里面就有pcm文件，我用的文件名为“16k-0.pcm”。例如https://ai.baidu.com/sdk#asr

4. 编译源代码，并执行。
	编译命令： g++ main.cpp -lwebsockets -lpthread -o aa
	执行： aa

#include <stdio.h>
#include <stdbool.h>
#include <string.h>
#include <stdlib.h>
#include <fcntl.h>
#include <iostream>
#include <mutex>
#include <thread>
#include <libwebsockets.h>

struct lws *websocket;
using namespace std;
void *readFileToMem(string file, size_t *read_size)
{
	FILE *fp;
	size_t file_size;
	void *mem = 0;

	fp = fopen(file.c_str(), "rb");
	if (fp == NULL) {
		printf("failed to open input file %s\n", file.c_str());
		*read_size = 0;
		return 0;
	}

	fseek(fp, 0L, SEEK_END);
	file_size = ftell(fp);
	rewind(fp);

	if (file_size <= 0)
	{
		fclose(fp);
		*read_size = 0;
		return 0;
	}

	mem = malloc(file_size);
	*read_size = fread(mem, 1, file_size, fp);
	if (*read_size != file_size)
	{
		free(mem);
		*read_size = 0;
		mem = 0;
	}

	fclose(fp);
	return mem;
}

int send_websockets(const void *data, int size, enum lws_write_protocol protocol)
{
	int ret;
	unsigned char *buf;

	buf = (unsigned char *)malloc(LWS_PRE + size);
	memcpy(&buf[LWS_PRE], data, size);	
	ret = lws_write(websocket, &buf[LWS_PRE], size, protocol);
	free(buf);
	return ret;
}

void send_start_params(void)
{
	int ret; // cuid值不能有空格
	static string req = R"({"type": "START", "data": {"appid": 42720845, "appkey": "bM1hT2tQHsFKmMqhYHRInGYC", "dev_pid": 1537, "cuid": "ygt_x5_tuyou", "sample": 16000, "format": "pcm"}})";

	ret = send_websockets(req.c_str(), req.size(), LWS_WRITE_TEXT);
	printf("[send_start_params]: size=%d\n", ret);
}

void send_audio(void)
{
	void *pcm;
	size_t size;
	int chunk_ms = 160;  // 160ms的录音
	int chunk_len = 16000 * 2 / 1000 * chunk_ms;

	pcm = readFileToMem("16k-0.pcm", &size);
	printf("pcm size = %d\n", size);
	
	for(int i = 0; i < size; i += chunk_len)
	{
		int ret;
		int send_size;
		
		if(i + chunk_len < size)
			send_size = chunk_len;
		else
			send_size = size - i;
		
		ret = send_websockets((unsigned char*)pcm + i, send_size, LWS_WRITE_BINARY);
		printf("[send_audio]: size=%d\n", ret);
		usleep(chunk_ms * 1000);
	}
}

void send_finish()
{
	int ret;
	static string req = R"({"type": "FINISH"})";
	
	ret = send_websockets(req.c_str(), req.size(), LWS_WRITE_TEXT);
	printf("[send_finish]: size=%d\n", ret);
}

void run(void)
{
	send_start_params();
	send_audio();
	send_finish();
}
 
static int callback_websocket(struct lws *wsi, enum lws_callback_reasons reason, void *user, void *in, size_t len)
{
    switch (reason) {
        case LWS_CALLBACK_CLIENT_CONNECTION_ERROR:
            fprintf(stderr, "Error connecting to server\n");
            break;
        case LWS_CALLBACK_CLIENT_ESTABLISHED:
		    {		    	
		    	fprintf(stderr, "Connected to server\n");
		    	thread th1(run);
		    	th1.detach();
		    }
            break;
        case LWS_CALLBACK_CLIENT_RECEIVE:
            fprintf(stderr, "Received data: %s\n", (char *)in);
            break;
        case LWS_CALLBACK_CLOSED:
            fprintf(stderr, "Disconnected from server\n");
            break;
        default:
            break;
    }
    return 0;
}

struct lws_protocols protocols[] = {
    { "default", callback_websocket, 65536, 65536 },
    { NULL, NULL, 0, 0 }
};

int main(int argc, char *argv[])
{
	struct lws_context_creation_info info = {};
	struct lws_client_connect_info i = {};
	struct lws_context *context;	
	
	info.port = CONTEXT_PORT_NO_LISTEN;
	info.protocols = protocols;
	info.gid = -1;
	info.uid = -1;
	context = lws_create_context(&info);
	if (context == NULL) {
		fprintf(stderr, "Creating libwebsocket context failed\n");
		return 0;
	}

	i.port = 80;
	i.path = "/realtime_asr?sn=123456xx";
	i.context = context;
	i.address = "vop.baidu.com";
	i.host = i.address;
	i.origin = i.address;
	websocket = lws_client_connect_via_info(&i);
    if (websocket == NULL) {
        fprintf(stderr, "Error connecting to server\n");
        return 0;
    }

    while(1)
    {
    	int ret;

    	ret = lws_service(context, 10000);  // 10ms
    	if(ret < 0)
    		break;
    }
    lws_context_destroy(context);
    printf("out\n");
    
	return 0;
}