Voice Capture DSP

An object that encapsulates several DSPs related to voice capture.

CLSID

CLSID_CWMAudioAEC

Interfaces

• IMediaObject
• IPropertyStore

Properties

Remarks

Unlike the other DSPs, the voice capture object encapsulates multiple DSPs in a single object, and the object is a DMO object only (it does not implement IMFTransform). The voice capture DMO includes the following DSP components:

• Acoustic echo cancellation (AEC)
• Microphone array processing
• Noise suppression
• Automatic gain control
• Voice activity detection

Applications can turn each component on and off individually.

The voice capture DMO supports two modes of operation, filter mode and source mode. In filter mode, the application sends audio samples from the microphone and from the speaker line to the DMO, and the DMO produces output.

In source mode, the application does not need to deliver samples to the DMO. Instead, the DMO manages all of the operations on the audio devices, including initializing the devices, capturing and synchronizing the audio streams, calculating time stamps, and retrieving the geometry of the microphone array. Using source mode, the application simply configures the DMO, and the output from the DMO is a clean, processed microphone signal. Source mode is significantly easier to use than filter mode, and is recommended for most applications.

Currently the voice capture DMO supports only single-channel acoustic echo cancellation (AEC), so the output from the speaker line must be single-channel. If microphone array processing is disabled, multi-channel input is folded down to one channel for AEC processing. If both microphone array processing and AEC processing are enabled, AEC is performed on each microphone element before microphone array processing.

Microphone Array Processing

A microphone array is a set of closely positioned microphones. Microphone arrays achieve better directionality than a single microphone, because the acoustic waves arrive at each microphone at a slightly different time. For more information on microphone arrays see the web articles Microphone Array Support in Windows Vista and How to Build and Use Microphone Arrays for Windows Vista.

Using the Voice Capture DSP

To use the Voice Capture DSP, perform the following steps.

1. Initialize the DMO

Create the voice capture DMO by calling CoCreateInstance with the CLSID CLSID_CWMAudioAEC. The voice capture DSDP exposes only the IMediaObject and IPropertyStore interfaces, so it can only be used as a DMO.

The DMO defaults to source mode. To select filter mode, set the MFPKEY_WMAAECMA_DMO_SOURCE_MODE property to VARIANT_FALSE.

Next, configure the internal properties of the DMO by using the IPropertyStore interface. The only property that an application must set is the MFPKEY_WMAAECMA_SYSTEM_MODE property. This property configures the processing pipeline within the DMO. The other properties are optional.

2. Set the Input and Output Formats

If you are using the DMO in filter mode, set the input format by calling IMediaObject::SetInputType. The input format can be almost any valid uncompressed PCM or IEEE floating-point audio type. If the input format does not match the output format, the DMO automatically performs sample-rate conversion.

If you are using the DMO in source mode, do not set the input format. The DMO automatically configures the input format based on the audio devices.

In either mode, set the output format by calling IMediaObject::SetOutputType. The DMO can accept the following output formats:

• Subtype: MEDIASUBTYPE_PCM or MEDIASUBTYPE_IEEE_FLOAT
• Format block: WAVEFORMAT or WAVEFORMATEX
• Samples per second: 8,000; 11,025; 16,000; or 22,050
• Channels: 1 for AEC-only mode, 2 or 4 for microphone array processing
• Bits per sample: 16

The following code sets the output type to 16-bit single-channel PCM audio:

DMO_MEDIA_TYPE mt;  // Media type.
mt.majortype = MEDIATYPE_Audio;
mt.subtype = MEDIASUBTYPE_PCM;
mt.lSampleSize = 0;
mt.bFixedSizeSamples = TRUE;
mt.bTemporalCompression = FALSE;
mt.formattype = FORMAT_WaveFormatEx;

// Allocate the format block to hold the WAVEFORMATEX structure.
hr = MoInitMediaType(&mt, sizeof(WAVEFORMATEX));
if (SUCCEEDED(hr))
{
    WAVEFORMATEX *pwav = (WAVEFORMATEX*)mt.pbFormat;
    pwav->wFormatTag = WAVE_FORMAT_PCM;
    pwav->nChannels = 1;
    pwav->nSamplesPerSec = 16000;
    pwav->nAvgBytesPerSec = 32000;
    pwav->nBlockAlign = 2;
    pwav->wBitsPerSample = 16;
    pwav->cbSize = 0;

    // Set the output type.
    if (SUCCEEDED(hr))
    {
        hr = pDMO->SetOutputType(0, &mt, 0); 
    }
    // Free the format block.
    MoFreeMediaType(&mt);
}

3. Process Data

Before processing any data, it is recommended to call IMediaObject::AllocateStreamingResources. This method allocates the resources used internally by the DMO. Call AllocateStreamingResources after the steps listed previously, not before. If you do not call this method, the DMO automatically allocates resources when data processing starts.

If you are using the DMO in filter mode, you must pass input data to the DMO by calling IMediaObject::ProcessInput. The audio data from the microphone goes to stream 0, and the audio data from the speaker line goes to stream 1. If you are using the DMO in source mode, you do not need to call ProcessInput.

To get output data from the DSP, perform the following steps:

1. Create a buffer object to hold the output data. The buffer object must implement the IMediaBuffer interface. The size of the buffer depends on the requirements of your application. Allocating a larger buffer can reduce the chances of glitches occurring.
2. Declare a DMO_OUTPUT_DATA_BUFFER structure and set the pBuffer member to point to your buffer object.
3. Pass the DMO_OUTPUT_DATA_BUFFER structure to the IMediaObject::ProcessOutput method.
4. Continue to call this method for as long as the DMO has output data. The DSP signals that it has more output by setting the DMO_OUTPUT_DATA_BUFFERF_INCOMPLETE flag in the dwStatus member of the DMO_OUTPUT_DATA_BUFFER structure.

Requirements

See also

Digital Signal Processors