来自实时源的HTML5音频流

Question

I am looking into implementing live audio streaming from a vxWorks system into an HTML5 audio player. I have been researching for a while but am stuck on one key step.

Work Flow:

1. A host has a mic and speaks into the mic
2. Audio is received into the vxWorks OS. Here it can be encoded, packaged - anything is possible here
3. ????
4. User opens web page to listen to the live audio in an HTML5 player.

I don't know what goes in step 3. Suppose I have now encoded the audio into Mp3. What technologies do I need to send this to the browser? I believe I can send this through HTTP procotol, but I am not understanding how that is packaged. Ie, how is audio packaged into an HTTP protocol. And what does the HTML5 player want as a source for this data. A URL? Or websocket data?

Thanks.

Answer 1

The solution for #3 depends on whether you need a low-latency live audio stream at the HTML5 player, or latency can be 10-30 seconds.

1. Latency can be high

You need a streaming server / HLS packager that will stream via HLS, and a webpage that hosts Flowplayer or JWPlayer which will play that HLS stream using HTML5 video tag. AAC-encoded (not mp3!) audio stream needs to be pushed (RTMP publishing is a most popular method) to that streaming server / HLS packager.

You could go with free nginx; it is portable to vxWorks and can stream out via HLS. You could also try free VLC or ffmpeg for the same.

If you make it work, the stream will be playable on any browser on any device - iOS, Android, Windows OS etc...

2. You need low latency

This is much harder. You would need another machine running Linux or Windows OS. On that machine install Unreal Media Server or Evostream server - these servers stream to HTML5 players via Websocket protocol using ISO BMFF packaging. Same as before, AAC-encoded (not mp3!) audio stream needs to be pushed (RTMP publishing is a most popular method) to that streaming server.

These streams will play on any browser / any OS EXCEPT iOS!