• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 99 subscribers
  • Views 784 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

G.729/iLBC/G.722: speech codec question

TIDR
Genius 9355 points

From a previous discussion I gather that  the current codec user guide is incorrect when it says XDM_RESET is not supported on the speech codecs. So, my customer is now trying to use XDM_RESET with G.729 speech decoder.

They tried doing XDM_RESET on the decoder in the middle of processing an audio stream, that is between two encoded frames, then it  takes several frames before the decoder will produce a good output. So, is this the way to handle packet losses or should they do something else when dealing with that?

By the way, the XDM_REset documentation issue will need to be fixed in G.729, G.722 and iILBC user guides.

Rgds

-Dipa-

  • 0
    Intellectual 315 points

    Dipa,

    "They tried doing XDM_RESET on the decoder in the middle of processing an audio stream, that is between two encoded frames, then it  takes several frames before the decoder will produce a good output"

    This is the expected behavior when we reset (XDM_RESET) the state of the codec. However, this is not the way to handle the packet losses at decoder. The decoder Inargs->frameType flag indicates whether the input frame is lost (ISPHDEC1_FTYPE_SPEECHLOST) or received correctly (ISPHDEC1_FTYPE_SPEECHGOOD). So, we need to set this flag to ISPHDEC1_FTYPE_SPEECHLOST for packet losses.

    "By the way, the XDM_REset documentation issue will need to be fixed in G.729, G.722 and iILBC user guides"

    We will update the documents for XDM_RESET issue in the next releases.

    -Venkat

  • 0 in reply to Venkateswara Rao
    TI__Genius 9355 points

    Venkat,

    Are all the frametypes outlined supported?

    typedef enum {

        ISPHDEC1_FTYPE_SPEECHGOOD = 0,  /**< Regular speech frame received

                                         *   without errors/loss.

                                         */

        ISPHDEC1_FTYPE_SIDUPDATE = 1,   /**< SID update frames. */

        ISPHDEC1_FTYPE_NODATA = 2,      /**< Untransmitted frame for codecs which

                                         *   support internal DTX.

                                         */

        ISPHDEC1_FTYPE_SPEECHLOST = 3,  /**< Complete loss of speech frame. */

        ISPHDEC1_FTYPE_DEGRADED = 4,    /**< Speech frame with a correct CRC,

                                         *   some invalid data.

                                         */

        ISPHDEC1_FTYPE_BAD = 5,         /**< Speech frame (likely), but invalid. */

        ISPHDEC1_FTYPE_SIDFIRST = 6,    /**< The first frame of comfort noise. */

        ISPHDEC1_FTYPE_SIDBAD = 7,      /**< Corrupt SID update frame. */

        ISPHDEC1_FTYPE_ONSET = 8        /**< Frames which precede the first speech

                                         *   frame of a speech burst.

                                         */

    } ISPHDEC1_FrameType;

     

    In the test app I only see the first 4 mentioned. 

    Also, what is SIDUPDATE vs SIDFIRST? How is each one to be used?

     

    Rgds

    -dipa-

  • 0 in reply to TIDR
    Intellectual 890 points

    Dipa,

    This is the enumeration defined for frametype, however, each codec can use only subset of these or all.

    Most of the codecs use only the first 4 values, GSM-AMR uses all the frametypes.

    SIDFIRST and SIDUPDATE indicate that the input frame is a silence. If it is the first frame of silence, frametype is set to SIDFIRST otherwise to SIDUPDATE.

    -Venkat