For the purposes of the present document, the following terms and definitions apply. H.324 terminal: 3G-324M terminal:

For the purposes of the present document, the following abbreviations apply: ACELP ADC AEC AL CCSRL CELP CT DAC DCT EI EOB FEC GOB GQUANT GTT HDLC HEC ISDN LAPM LC MC MCU MP-MLQ MPL MR-ACELP PC MCU MUX PDU SN VLC

For a video telephone this would most likely consist of a video camera and display monitor. Other possible input sources could be a VCR or disk drive. While most applicable I/O equipment relies on a standard format for the video signal or bit stream, this format is likely to differ from that mandated by the video codec. In such cases, circuitry or software is used to transcode between the two formats.

ITU-R 601 (NTSC or PAL) is a typical video input signal and represents a bit stream of 20.7 Mbyte/s for the actual image (excluding blanking intervals). The first order of compression occurs by reducing the resolution of the input signal. For example, CIF resolution at 30 fps produces a bit stream 4.6 Mbyte/s. Additional savings occur by dropping frames. In a videoconference, where motion is relatively slow, 10 fps is considered adequate. Thus, the original signal of 20.7 Mbytes/s could be reduced to 1.5 Mbyte/s with just these techniques. However, this is still 188 times greater than can be transmitted on, for example, a 64 KBPS channel. Substantial compression is still required, especially considering that framing, control, and audio would as well require a portion of the available bandwidth. To achieve the degree of compression required for video telephony, all of the video codecs that can currently be employed in a 3GPP multimedia codec use a combination of spatial and temporal redundancy reduction to reduce the bandwidth required by the video media stream. Spatial redundancy can be reduced by converting the input signal from the time domain to the frequency domain using a DCT. This produces a DC value and other coefficients, where most of the scene energy is concentrated in the coefficients corresponding to the lower frequencies. Next, a coarse quantizer is applied (which, in this domain, has little effect on image quality). This results in many of the coefficients being encoded to 0. The significant coefficients are encoded to a much smaller range of values. The coefficients are then reordered so that, typically, the larger magnitude values will occur first followed by 0 value coefficients. Finally, the coefficients are replaced with a count of the number of zero value coefficients followed by the value of a nonzero coefficient. This combination is translated into a VLC. Applying this type of compression to the entire video frame produces an intra frame. Despite the efficiency of intra coding, significantly more compression is required. In addition to removing spatial redundancy, all video codecs apply temporal reduction as well. This is achieved by comparing the current frame to the previous and estimating the set of vectors which when applied to their respective areas of the scene would create the new, current frame based on the old, previous frame. The match is usually not perfect, so an error component is transmitted as well. The error component is also transformed to the frequency domain, so the same compression efficiency achieved in the intra frame is achieved here as well - enhanced by the fact the range of error coefficients is less than intra coefficients. Since generally only a few areas of a scene change from frame to frame, high compression can be achieved by sending a series of inter frames. If the error component for a particular block is too large, it can be encoded as an intra block. Since, by their nature, VLCs are not fixed length, a single bit error can make it impossible to decode an entire frame. Unfortunately, each inter coded frame relies on its previous frame to be decoded. Thus, a single bit error can destroy the entire remaining bit stream. Video codecs have various ways of handling errors. The simplest is to use error detection to determine if a frame contains an error. The transmitter is then signalled that an error occurred. It then sends an intra coded frame, which does not depend on any previous frames. This approach consumes considerable bandwidth and is only practical for very low error networks. Other, more sophisticated schemes are available using the video codecs available to 3G-324M terminals.

Generally, a video telephone would require a handset, headset, or microphone and speaker. Often, integrated circuits are employed that convert the typically analogue input signal to a PCM format bit stream (ADC) and convert PCM to an analogue signal for acoustic output (DAC). This is helpful since many speech codecs use PCM for input and output. Video telephones often use a separate microphone and speaker. This allows the user to be seen without a handset or headset. However, if this is so, AEC will be required.

Various data protocols can be supported. These always support data applications (see clause A.5 User Data Applications). A specific protocol or set of protocols is often stipulated by the data application. Each protocol provides varying degrees of error detection and/or correction.

In general, system control constitutes the overall state machine for the terminal. It usually has to be aware of when a connection has been established. At that point it can begin H.245 procedures such as master/slave determination, capabilities exchange, and opening logical channels. Upon call termination, either initiated at the near or far ends, system control generally initiates H.245 end session procedures.

All out-of-band network signalling for the purpose of call control is handled by call set-up, which is usually implemented as a state machine. This includes initiating, answering, and tearing down calls.

H.245 specifies the syntax and semantics for in-band, terminal-to-terminal control messages and the procedures for their use. Most importantly, H.245 is used for master/slave determination, capabilities exchange, H.223 mux table transmission, and opening and closing logical channels. There is also a large array of general control and indication messages. H.245 addresses a wide range of terminals and applications. Therefore, only a subset of the messages listed in H.245 pertain to 3G-324M terminals. Messages fall into one of four categories: Request (requires a Response), Response (in response to a Request), Command (requires an action), and Indication (informative only). H.245 messages are carried on a single logical channel within the H.223 mux. This channel is labelled LC 0 and is considered to be open upon establishing digital communications end-to-end and survives until digital communication is terminated. Due to the characteristics of the H.223 mux, bandwidth for H.245 messages is allocated on an as-needed basis. Since most H.245 traffic occurs and the beginning and end of the session, this conserves much needed bandwidth for video and audio. Error control is not provided for within H.245 and is specified elsewhere.

H.223 describes the multiplexing protocol used between H.324 terminals. It is packet oriented and each packet can contain a subset of a maximum of 65536 LCs. Each LC represents a single media, information, or control channel. The H.223 protocol is split into two layers, the lowest being the Multiplex Layer. The Multiplex Layer exchanges data with the end terminal via MUX-PDUs. Multiplex table entries, of which there are 16 (and can be changed during a session), describe which octets from within the PDU are allocated to which logical channels. The multiplex table entry employed for a particular PDU is indicated by the 4 bit MC field in the MUX-PDU header. MUX-PDUs contain an integer number of octets. Errors within the MUX-PDU header are controlled using the HEC field in the MUX-PDU header. H.324 terminals utilising the V.34 transmission protocol frame MUX-PDUs with HDLC. Bit stuffing is used for data transparency in this case. Above the Multiplex Layer is the Adaptation Layer, of which there are three different types.

1. AL1 is designed primarily for control information and data protocols. It can be either framed or unframed and does not provide any error control.
2. AL2 is designed primarily for the transfer of digital audio. AL2 PDUs contain 1 octet for an 8-bit CRC and an optional octet for a sequence number.
3. AL3 is designed primarily for the transfer of digital video. AL3 PDUs contain 2 octets for a 16-bit CRC. There is also optionally 1 or 2 octets for control. AL3 also allows limited retransmission.

For purposes of video telecommunications over wireless networks, four annexes to H.223 have been created. These create four levels of error detection and error correction.