adiCommandRecord

Sends a data array containing raw commands to an actively running recording function. Use adiCommandRecord to enable and configure voice activity detection.

Supported board types

• AG

• CG

Prototype

DWORD adiCommandRecord ( CTAHD ctahd, WORD *data [], DWORD nwords)

Return values

Events

Note: The application receives ADIEVN_RECORD_EVENT asynchronously, while the speech buffers arrive every buffersize x framerate / framesize msec, attached to ADIEVN_RECORD_BUFFER_FULL (when speech is detected).

Details

The following DSP file must be loaded to the board to enable voice activity detection:

To configure CG boards for voice activity detection, specify rvoice_vad in the resource definition. For example:

Resource[0].Definitions = (dtmf.det_all & rvoice_vad.rec_alaw & rvoice_vad.play_alaw...

To enable voice activity detection, call adiCommandRecord on an actively running ADI recording function (such as adiRecordAsync). Automatic gain control and energy detection must be disabled when using voice activity detection. Recording must be using ADI_ENCODE_MULAW, ADI_ENCODE_ALAW, or ADI_ENCODE_PCM8M16.

Call adiCommandRecord after receiving ADIEVN_RECORD_STARTED.

The first parameter must be one of the following voice activity detector commands:

When enabling voice activity detection (EVAD_CDE_FUNCTION_ENABLE), modify the voice activity detector's default behavior with the following parameters (also defined in /nms/include/evad.h):

Convert signal_attack, signal_release, noise_attack, and noise_release into DSP format using the following formula:

/* time constant (tc) and period need to have the same units of time */
int epsilon(float tc,float period)
{
float eps;
eps =  (float)(1.0 - (float) exp((double)((-1.0 * period)/ tc)));

return (int) (eps * 32767);   // return in S.15 format
}
LATENCY = 10; /* 10 msec record DPF period */
DSP_value = epsilon ( time_constant, LATENCY );

The custom recording DPF sends data to the host by calling the DSPOS function dspkSendEvent. The first parameter sent by the function displays in the value field of the CTA_EVENT structure. All remaining parameters display in an attached buffer. The application is responsible for freeing the buffer after it processes the data.

For more information, refer to Detecting voice activity.

Example

   /* This code sends a command to custom record DPF and prints the events from the DPF. */
myWaitForEvent( ctaqueuehd, event );
switch(event->id)
{
    /* etc... */
    case ADIEVN_RECORD_STARTED:
        {
               WORD myParms[3] = { 0x1111, 0x2222, 0x3333 );
               adiCommandRecord( ctahd, myParms, 3);
               /* At the DSP level, the DPF will see the following
                *  command packet.
                *
                * 0x3        -> size of command packet
                * 0x1111     -> parm 1
                * 0x2222     -> parm 2
                * 0x3333     -> parm 3
                */
        }
    case ADIEVN_RECORD_EVENT:
        if (event->buffer != NULL) // event with multiple data
        {
            WORD i;
            WORD *pData = (WORD *) event->buffer;
            printf("event->value  %x\n", event->value );
            printf("event->size  %x\n", event->size );
            for (i=0; i < event->size / sizeof(WORD); i++)
            {
                printf("data[%d] %x\n", i, pData[i]);
            }
            if (event->size & CTA_INTERNAL_BUFFER)
            {
                ctaFreeBuffer( event->buffer );
                printf("Buffer freed\n");
            }
        }
        else
        {                          // event with only 1 data
            printf("event->value %x\n", event->value );
        }
        break;
}