基於EasyDarwin 流媒體播放器的客戶端開發
最近接觸了一個國內優秀的流媒體平臺- EasyDarwin。這是一個國外開源流媒體平臺Darwin深度裁剪版本，看了一段時間後就想跟自己的開發的一個android裝置對接，以了卻我多年對多媒體更深入的理解。（本人曾經自己開發一個H264的移動電視的軟解碼播放器）
在搭建好伺服器後，拿出了全志的T2開發板，開始使用其團隊的EasyPusher庫進行推送。首先我嘗試視訊直播的開發，在研究了兩天後，發現為我的機器的系統的解碼器已經被封裝成庫，對外不開放。所以我不得不使用其android系統的mediaRecoder進行編碼，而系統已經把其直接打包成了MP4檔案，這樣我無法獲得直接的H264資料。於是陷入了困頓之中。在等待方案公司的幫助的時候，我尋找了別的方法去獲得推送的資料流，最後不得不採用localhost技術，該技術就是將mediaRecoder編碼的資料通過類似本地網路管道
第一步：搜尋MDAT資料頭
第二步：將其中的資料解析成H264的資料包，其中資料幀的長度剛好四個位元組，用
第三步：呼叫EasyPusher將資料傳送到伺服器。這裡需要注意的是，遇到I幀的時候需要加上SPS和PPS。當然localhost的資料是不會有這些資料的，只好自己根據解析度等資訊寫好。如下
const unsigned char SPSPPSCIF[] ={0X00,0X00,0X00,0X01, 0x67,0x42, 0X00, 0X1F, 0XE5, 0X40, 0XA0, 0xfc,0X80,0X00,0X00,0X00,0X01,0X68,0XCE,0x31,0X12};//320*240
const unsigned char SPSPPS576[] ={0X00,0X00,0X00,0X01, 0x67,0x42, 0X00, 0X1F, 0XE5, 0X40, 0X5A, 0x09,0X32,0X00,0X00,0X00,0X01,0X68,0XCE,0x31,0X12}; //720*576
這些資料最好出現I幀(幀頭第5個位元組一般是0X67，參照H264資料規範)加上。我估計這對流媒體傳輸來說是很重要的。
傳輸過去後，在客戶端可以看到視訊，但丟幀非常嚴重。我估計主要是因為localhost以及編碼器效率太低，同時時間戳也是一個估計數。所以造成了丟幀現象。沒有什麼好的辦法了，只有等待硬體編碼器的相關資料直接編碼進行傳送了。
繞過直播，開始搞回放。首先選擇一個DEMUX工具，我選擇的是MP4V2，主要是這個工具WINDOWS和android NDK版本都有，所以相對來說比較省時間。折騰一段時間後發現數據能傳輸上去了。但是音訊和視訊完全不同步。在跟EasyDarwin 團隊的一個朋友討論後我決定開始搞音視訊同步，策略是採用的是講音視訊幀都同步到系統時間。果然效果還是很明顯的，除了偶爾會卡住基本上能流暢播放。我估計著需要開始瞭解播放器端的同步了。相關的程式碼附錄上：
MP4 DEMUX：獲取視訊和音訊幀：
typedef struct _Media_INFO_
{
MP4FileHandle mp4File;
MP4TrackId video_trId;
MP4TrackId audio_trId;
u_int32_t video_sample_max_size;
u_int32_t video_num_samples;
u_int32_t video_width;
u_int32_t video_height;
double video_frame_rate;
u_int32_t video_timescale;
u_int32_t audio_num_samples;
int audio_num_channels;
u_int32_t audio_timescale;
MP4Duration audio_duration;
u_int32_t avgBitRate;
u_int8_t AudioType;
u_int8_t *p_audio_config;
u_int32_t n_audio_config_size;
u_int32_t n_video_config_size;
u_int32_t audio_sample_max_size;
MP4Duration video_duration;
u_int8_t *p_Vediobuffer;
u_int8_t *p_Audio_buffer;
}MEDIA_INFO;
staticint GetVideoStreamHeader(MP4FileHandle mp4File, MP4TrackId video_trId, intvideo_codec,
unsigned char*strm_hdr_buf, int *strm_hdr_leng)
{
int b;
// for MPEG4
unsigned char *p_video_conf;
unsigned int n_video_conf;
// for H.264
unsigned char **pp_sps,**pp_pps;
unsigned int *pn_sps, *pn_pps;
unsigned int n_strm_size;
int i;
switch (video_codec) {
case RAW_STRM_TYPE_M4V: //MPEG4
p_video_conf = NULL;
n_video_conf = 0;
b = MP4GetTrackESConfiguration(mp4File, video_trId,
&p_video_conf, &n_video_conf);
if (!b)
return -1;
memcpy(strm_hdr_buf, p_video_conf, n_video_conf);
free(p_video_conf);
*strm_hdr_leng = n_video_conf;
break;
case RAW_STRM_TYPE_H263: //H.263
*strm_hdr_leng = 0;
break;
case RAW_STRM_TYPE_H264RAW: // H.264
pp_sps = pp_pps = NULL;
pn_sps = pn_pps = NULL;
n_strm_size = 0;
b = MP4GetTrackH264SeqPictHeaders(mp4File, video_trId, &pp_sps,&pn_sps, &pp_pps, &pn_pps);
if (!b)
return -1;
// SPS memcpy
if (pp_sps) {
for (i=0; *(pp_sps + i); i++) {
memcpy(strm_hdr_buf +n_strm_size, h264_delimiter, sizeof(h264_delimiter));
n_strm_size +=sizeof(h264_delimiter);
memcpy(strm_hdr_buf +n_strm_size, *(pp_sps + i), *(pn_sps + i));
/*
if(NAL_UNIT_TYPE_TYPE(strm_hdr_buf[n_strm_size]) == 7) {
strm_hdr_buf[n_strm_size +1] = 66;
}
*/
n_strm_size += *(pn_sps + i);
free(*(pp_sps + i));
}
free(pp_sps);
}
// PPS memcpy
if (pp_pps) {
for (i=0; *(pp_pps + i); i++) {
memcpy(strm_hdr_buf +n_strm_size, h264_delimiter, sizeof(h264_delimiter));
n_strm_size +=sizeof(h264_delimiter);
memcpy(strm_hdr_buf + n_strm_size,*(pp_pps + i), *(pn_pps + i));
n_strm_size += *(pn_pps + i);
free(*(pp_pps + i));
}
free(pp_pps);
}
*strm_hdr_leng = n_strm_size;
break;
default: // Unknown
*strm_hdr_leng = 0;
break;
}
return 0;
}
int GetAudioFrame(MEDIA_INFO* pMedia,unsigned int *pSize, intsample_id, unsigned long* timebase ,unsigned long* dutation)
{
int nRead;
u_int8_t *p_audio_sample;
u_int32_t n_audio_sample;
int b, ret;
MP4Timestamp TimeStamp;
MP4Duration Audio_duration;
p_audio_sample = (u_int8_t *) pMedia->p_Audio_buffer;;
*pSize= pMedia->audio_sample_max_size;
n_audio_sample = *pSize;
if ((sample_id <= 0) || (sample_id >pMedia->audio_num_samples)) {
*pSize = 0;
return -1;
}
/////////////////////
// MP4ReadSample //
/////////////////////
b= MP4ReadSample(pMedia->mp4File,pMedia->audio_trId, sample_id,
&p_audio_sample,&n_audio_sample,
&TimeStamp,&Audio_duration, NULL, NULL);
if (!b) {
*pSize = 0;
return -1;
}
*pSize = nRead = n_audio_sample;
*timebase= TimeStamp;
*dutation= Audio_duration;
return nRead;
}
int GetVideoFrame(MEDIA_INFO* pMedia, intm_video_codec, unsigned int *pSize,unsigned int *isIframe, int sample_id,unsigned long* timebase, unsigned long*duration )
{
int nRead;
int n_video_hdr;
u_int8_t *p_video_sample;
u_int32_t n_video_sample;
MP4Timestamp TimeStamp;
MP4Duration video_duration;
int b, ret;
p_video_sample = (u_int8_t *) pMedia->p_Vediobuffer;
*pSize= pMedia->video_sample_max_size;
n_video_sample = *pSize;
if((sample_id <= 0) || (sample_id > pMedia->video_num_samples)) {
*pSize = 0;
return -1;
}
*isIframe = MP4GetSampleSync(pMedia->mp4File,pMedia->video_trId, sample_id);
if (*isIframe ) {
ret = GetVideoStreamHeader(pMedia->mp4File, pMedia->video_trId,m_video_codec,
p_video_sample, &n_video_hdr);
p_video_sample += n_video_hdr;
}
else
n_video_hdr = 0;
/////////////////////
// MP4ReadSample //
/////////////////////
b= MP4ReadSample(pMedia->mp4File, pMedia->video_trId, sample_id,
&p_video_sample,&n_video_sample,
&TimeStamp,&video_duration, NULL, NULL);
if (!b) {
*pSize = 0;
return -1;
}
//if (codec == h.264), the first 4 bytes are the length octets.
// They need to be changed to H.264 delimiter (00 00 00 01).
if (m_video_codec == RAW_STRM_TYPE_H264RAW) {
int h264_nal_leng;
int nal_leng_acc = 0;
do {
h264_nal_leng = p_video_sample[0];
h264_nal_leng = (h264_nal_leng << 8) | p_video_sample[1];
h264_nal_leng = (h264_nal_leng << 8) | p_video_sample[2];
h264_nal_leng = (h264_nal_leng << 8) | p_video_sample[3];
memcpy(p_video_sample, h264_delimiter, 4);
nal_leng_acc += (h264_nal_leng +4);
p_video_sample += (h264_nal_leng + 4);
} while (nal_leng_acc < n_video_sample);
}
*timebase= TimeStamp;//TimeStamp+video_duration;
*duration= video_duration;
*pSize = nRead = (n_video_hdr + n_video_sample);
return nRead;
}
時鐘同步：
/返回需要延遲的時鐘
typedef struct _SYN_CLOCK_CTRL_
{
unsignedchar Vedioflag;
unsignedchar Audioflag;
unsignedlong ClockBase;
unsignedlong ClockCurr;
unsignedlong VedioBase;
unsignedlong AudioBase;
}Sync_clock_Ctl;
Sync_clock_Ctl g_clock;
unsigned long Sync_clock(MEDIA_INFO*pMedia, unsigned long timebase, unsigned long duration, int type, unsignedlong* OutTime)
{
unsignedlong TimebaseNew;
unsignedlong DiffClock;
doubleTimeCalbase;
doubleTimenext;
unsignedlong CurrentTime;
unsignedlong NextTime;
unsignedlong delay;
unsignedlong TimeScale;
#ifdef _WIN32
if(g_clock.ClockBase== 0)
{
g_clock.ClockBase= ::GetTickCount();
}
g_clock.ClockCurr= ::GetTickCount();
#else
{
structtimeval tv;
gettimeofday(&tv,NULL);
//return(int64_t)tv.tv_sec * 1000000 + tv.tv_usec;
g_clock.ClockCurr= (int64_t)tv.tv_sec * 1000 + tv.tv_usec/1000;
if(g_clock.ClockBase== 0)
{
g_clock.ClockBase= g_clock.ClockCurr;
}
}
#endif
DiffClock= g_clock.ClockCurr - g_clock.ClockBase;//時鐘的耗時間Tick數//微妙級別忽略不計
if(type== VEDIO_PUSH)
{
if(g_clock.Vedioflag== 0)
{
g_clock.VedioBase= timebase;
g_clock.Vedioflag= 1;
}
TimeScale= pMedia->video_timescale;
}else
{
if(g_clock.Audioflag== 0)
{
g_clock.AudioBase= timebase;
g_clock.Audioflag= 1;
}
TimeScale= pMedia->audio_timescale;
}
TimeCalbase= ((double)(timebase-g_clock.VedioBase))/TimeScale;
Timenext= ((double)(timebase-g_clock.VedioBase+duration))/TimeScale;
//開始計算當前和小一個Sample的時間估計決定延遲//
CurrentTime= (unsigned long)(TimeCalbase*1000);
NextTime= (unsigned long)(Timenext*1000);
*OutTime= CurrentTime;
if(DiffClock> NextTime) //已經落後，快進
{
delay= 0;
}else
{
delay= NextTime- DiffClock;//重新計算時間
}
returndelay;
}