MP design and implementation CS414: Multimedia System Instructor: Klara Nahrstedt April 13, 2012

Covered Aspects of Multimedia in CS414 MP Image/Video Capture Media Server Storage Transmission Compression Processing Audio/Video Presentation Playback Audio/Video Perception/ Playback Audio Information Representation Transmission Audio Capture A/V Playback Image/Video Information Representation

MP1: AV Capture and Playback Image/Video Capture Compression/ Processing Audio Information Representation Audio Capture Audio/Video Perception/ Playback

MP1: Video Capture and Store in a File Sr c Sink Src Sink Src Video Source Video Filter Media Encoder Sink Src Video Muxer Sink Video Sink final Element videosrc = ElementFactory.make("v4l2src", "source"); final Element videofilter = ElementFactory.make("capsfilter", "flt"); videofilter.setCaps(Caps.fromString("video/x-raw-yuv, width=640, height=480, framerate=10/1")); final Element formatConverter = ElementFactory.make("ffmpegcolorspace", "formatConverter"); final Element encoder = ElementFactory.make("ffenc_mpeg4", "encoder"); final Element muxer = ElementFactory.make("fmux_mpeg", "muxer"); final FileSink filesink = (FileSink) ElementFactory.make("filesink", "filesink"); filesink.setLocation("capture01.avi"); Snippet pipe.addMany(videosrc,formatConverter, videofilter, encoder, filesink); Element.linkMany(videosrc, formatConverter, videofilter, encoder, filesink); pipe.setState(State.PLAYING); Gst.main(); pipe.setState(State.NULL); START

Video Capture and Store in a Queue Video Source Video Filter final Element videosrc = ElementFactory.make("v4l2src", "source"); final Element videofilter = ElementFactory.make("capsfilter", "flt"); videofilter.setCaps(Caps.fromString(Caps.fromString(String.format("video/x-raw-yuv, width=%s, height=%s" + ", bpp=24, depth=16, framerate=%s/1",WIDTH+"", HEIGTH+"", CAPTURERATE+""))); videorate = ElementFactory.make("videorate", "videorate"); ratefilter = ElementFactory.make("capsfilter", “rateFilter"); ratefilter.setCaps(Caps.fromString(String.format("video/x-raw-yuv, framerate=%s/1”,FRAMERATE+""))); final Element encoder = ElementFactory.make(” theoraenc", ”H264"); final Element muxer = ElementFactory.make(”oggmux", "muxer"); final AppSink appsink = (FileSink) ElementFactory.make(”appsink", ”appsink"); Snippet Sr c Sink Src Media Encoder Video Muxer Sink Src Sink Src Sink App Sink Video Rate Rate Filter Sink Src Sink Src

Gstreamer-java AppSink Sink App Sink Element A callback function is used to get the frames from the AppSink Element appsink.set("emit-signals", true); appsink.setSync(false); appsink.connec(new AppSink.NEW_BUFFER(){ public void newBuffer(AppSink as){ Buffer buffer = as.getLastBuffer(); ByteBuffer buf = buffer.getByteBuffer(); byte[] b = new byte[buf.remaining()]; buf.get(b); int timestamp = (int)buffer.getTimestamp().toMillis(); ApplicationPacket packet = new ApplicationPacket(b, timestamp) Queue.enQueue(packet); } }); Queue Packet ArrayBlockingQueue

MP1: Video Playback from a File Pipeline pipe = new Pipeline(”Decode pipeline"); Element videosrc = ElementFactory.make("filesrc", ”filesrc"); src.set("location", "Avatar.avi"); DecodeBin2 decodeBin = new DecodeBin2("Decode Bin"); Element decodeQueue = ElementFactory.make("queue", "Decode Queue"); final VideoComponent videoComponent = new VideoComponent(); pipe.addMany(videosrc, decodeQueue, decodeBin); Element.linkMany(videosrc, decodeQueue, decodeBin); decodeBin.connect(new DecodeBin2.NEW_DECODE_PAD()){ public void newDecodePad (DecodeBin2 elem, Pad pad, boolean last){ Caps cap = pad.getCaps(); Structure struct = caps.getStructure(0); if(struct.getName().startsWith(“video/”){ pad.link(videoComponent.getElement.getStaticPad(“Sink”)); }}} Sr c Sink Src Media SourceMedia Decoder Sink Src Media DeMuxer Sink Media Sink Sink Src DecodeBin

Video Playback from a Queue Sr c Sink Src Sink Src App Source Video Filter Media Encoder Sink Src Video Muxer Sink Video Sink final AppSrc appsrc = (AppSrc) ElementFactory.make("appsrc", ”appsrc"); appsrc.setCaps(Caps.fromString(Caps.fromString(String.format("video/x-raw-yuv, width=%s, height=%s" + ", bpp=24, depth=16, framerate=%s/1",WIDTH+"", HEIGTH+"", CAPTURERATE+""))); final Element decodeBin = ElementFactory.make("decodebin2","decode"); final VideoComponent videoComponent = new VideoComponent(); pipe.addMany(appsrc, decodeBin); Element.linkMany(appsrc, decodeBin); decodeBin.connect(new DecodeBin2.NEW_DECODE_PAD()){ public void newDecodePad (DecodeBin2 elem, Pad pad, boolean last){ Caps cap = pad.getCaps(); Structure struct = caps.getStructure(0); if(struct.getName().startsWith(“video/”){ pad.link(videoComponent.getElement.getStaticPad(“Sink”)); } Snippet Sink Src Decode Bin

Gstreamer-java AppSrc Source App Source Element A callback function is used to put the frames into the AppSrc Element appsrc.set("emit-signals", true); appsrc.setSync(false); appsrc.connect(new AppSrc.NEED_DATA() { public void needData(AppSrc as, int size) { ApplicationPacket packet = Queue.dequeue(); while (packet == null) { System.out.println("Null data. trying again"); packet = Queue.dequeue(); } byte[] data = packet.getData(); Buffer buffer = new Buffer(data.length); buffer.getByteBuffer().put(data); as.pushBuffer(buffer); } {); Queue ArrayBlockingQueue Packet

MP2 Design Space Image/Video Capture Media Server Storage Transmission Compression Processing Audio/Video Presentation Playback Audio/Video Perception/ Playback Audio Information Representation Transmission Audio Capture A/V Playback Image/Video Information Representation

MP2: Audio Video Streaming Server Client Server –Capture Audio and Video at a fixed rate –Video: 30 fps, Audio: 8000Hz Client –Requests for Video and/or Audio –Works in Two modes: Active Mode and Passive Mode Control Channel Data Channel Active Mode: Media type: Audio, Video Video Rate: 15 to 25 fps Audio Rate: 8000Hz Video Resolution: 640X480 Passive Mode: Media type: Video Video Rate: 10 fps Video Resolution: 320X240

Video Data Flow Architecture: Server RTP Packet Rate Control Network Video SourceVideo Filter Media EncoderVideo Muxer App Sink Payload Media Type Timestamp RTP header ID UDP Packet Payload Segment

UDP Communication: Server Payload Media Type Timestamp RTP Packet ID UDP Packet Payloads PID SEQ SIZE PAYLOAD SIZE UDP Packet Payload System.arraycopy(pid_byte, 0, senddata, 0, pid_byte.length); System.arraycopy(seq_byte, 0, senddata, 4, seq_byte.length); System.arraycopy(size_byte, 0, senddata, 8, size_byte.length); While(true){ int remaining = (databyte.length - from); if(remaining < 0 ) break; else if(remaining > UDP_PAYLOAD_SIZE){ byte[] payload_size_byte = toBytes(UDP_PAYLOAD_SIZE); System.arraycopy(payload_size_byte, 0, senddata, 12, payload_size_byte.length); System.arraycopy(databyte, from, senddata, 16, UDP_PAYLOAD_SIZE); } else{ byte[] payload_size_byte = toBytes(remaining); System.arraycopy(payload_size_byte, 0, senddata, 12, payload_size_byte.length); System.arraycopy(databyte, from, senddata, 16, remaining); } sendData(7777, ”IP_Address", senddata); seq++; from += UDP_PACKET_SIZE; DatagramSocket DatagramPacket UDP Packet Header

Video Data Flow Architecture: Client Network RTP Packet App Source Decode Bin2 Video Sink Synchronization Payload Media Type Timestamp RTP header ID UDP Packet Payload Merge

UDP Communication: Client Payload Media Type Timestamp RTP Packet ID PID SEQ SIZE PAYLOAD SIZE UDP Packet Payload System.arraycopy(recv_data, 0, pid_byte, 0, 4); System.arraycopy(recv_data, 4, seq_byte, 0, 4); System.arraycopy(recv_data, 8, size_byte, 0, 4); System.arraycopy(recv_data, 12, payload_size_byte, 0, 4); HashTable index with PID SIZE if(!received_frame.containsKey(pid)){ ApplicationPacket app_frame = new ApplicationPacket(); app_frame.frame = new byte[size]; app_frame.array_size = size; System.arraycopy(recv_data, 16, app_frame.frame, seq*UDP_PACKET_SIZE, payload_size); received_frame.put(pid, app_frame); app_frame.array_size -= payload_size; }else{ ApplicationFrame app_frame = (ApplicationFrame) received_frame.get(pid); System.arraycopy(recv_data, 16, app_frame.frame, seq*UDP_PACKET_SIZE, payload_size); app_frame.array_size -= payload_size; received_frame.put(pid, app_frame); } if(app_frame.array_size == 0) finish_frame(pid); received_frame

Video Data Flow Architecture RTP Packet Rate Control Network Video SourceVideo Filter Media EncoderVideo Muxer App Sink RTP Packet App Source Decode Bin2 Video Sink Synchronization Payload Media Type Timestamp RTP header ID UDP Packet Payload Segment Merge

Rate Control: QoS Enforcement You can build leaky bucket You can implement token bucket You can use gstreamer rate control You can implement your own method Sleep (100) => 10 fps You can simply read using a loop While(true){ SendPacket(); Thread.sleep(10 0); }

Resource Admission: Client Client –Available Application Bandwidth AB N –Application Frame Size, M N = ? –Application Frame Rate, R N = ? –Audio Bandwidth = 8000 * 16 –Request Bandwidth Active Mode: B N = (M N * R N ) + Audio Bandwidth Passive Mode: B N = (M N * R N ) resource.txt Optimistic Allocation 8000Hz Audio Signal How to define R N for active mode? Show your computation on the Screen fps: 10

Resource Admission: Client Client –Available Application Bandwidth AB N –Application Frame Size, M N = ? –Application Frame Rate, R N = ? –Request Bandwidth Active Mode: B N = (M N * R N ) + Audio Bandwidth resource.txt Optimistic Allocation Show your computation on the Screen AB N = (M N *R N ) + Audio Bandwidth if R N > 25, R N = 25 if R N < 15, REJECT else R N

Resource Admission: Client Server –Available Application Bandwidth AB N –Used Bandwidth for Server 1 B 1 –Available Bandwidth AB N = AB N –B 1 –Admission is successful for B 2 if B 2 <= AB N resource.txt Server 160 Kbps Server 235 KBps Resource Table

Using Server Video SourceVideo Filter Tee Video ScaleScale Filter Video Rate Rate Filter Video Encoder RTP Payload RTP Bin UDP Sink Queue UDP Sink UDP Source RTCP RtpBin combines RtpSession RtpSsrcDemux RtpPtDesmux RtpJitterBuffer

Using Client Video SinkVideo DecoderRTP DePayload RTP Bin UDP Source UDP Sink RTCP rtpBin.connect(new Element.PAD_ADDED() { public void padAdded(Element element, Pad pad) { if (pad.getCaps().toString().contains("video")) { //LINK TO VIDEO SINK } else if (pad.getCaps().toString().contains("audio")) { // LINK TO AUDIO SINK} } });

Internals of RtpBin l

MP3: Multi-view Surveillance System Window for server 1 Window for server 2 Server –Capture Audio and Video at a fixed rate –Video: 30 fps, Audio: 8000Hz Client –Requests for Video and/or Audio –Works in Two modes: Active Mode and Passive Mode Active Mode: Media type: Audio, Video Video Rate: 15 to 25 fps Audio Rate: 8000Hz Video Resolution: 640X480 Passive Mode: Media type: Video Video Rate: 10 fps Video Resolution: 320X240

Additional Features for MP3? It is highly recommended that you build an additional feature in MP3 to SURPRISE US! Bonus Point: 20 Bonus point will be awarded based on the amount of surprise!