Spectrum Averaging with Hypersignal RIDE

Spectrum averaging is performed for a variety of reasons in DSP and related engineering fields. The averaging of a spectrum in time allows for the relative contribution of ambient noise to be effectively identified, and a noise characteristic obtained. Additionally, averaging may be used in a variety of recognition algorithms to identify background noise environments and distinguish certain sounds from other sounds. There exists several methods of achieving this averaging of a waveform and it need not apply purely to spectrum processing – virtually any signal which can benefit from a frame-based average can employ the following techniques.

Detailed Description

This application note will demonstrate averaging using three different techniques, using the Global Frame Mean function, the Power Spectrum function, and the Running Frame Combiner function.

Averaging using the Global Frame Mean Function

In this method, a frame-based running sum is calculated and divided by a running frame count, effectively giving the dynamic frame-based average over time. The output waveform, y[m,n], is given as follows:
The second will make use of the Power Spectrum function which employs an exponential filter in time for the averaging. The output waveform, y[m,n], is given as follows:
The final method will employ a Running Frame Combiner to calculate the average frame based on the current average and the current incoming frame. In this case, the average is given by:
In the Hypersignal RIDE application, as shown in Figure 1, spectrum averaging from a time-domain disk file can be accomplished (after performing an FFT and magnitude function) with Global Frame Mean block (found in Statistical Functions group) as follows:
In the Hypersignal RIDE application, as shown in Figure 2, frame-based averaging from a disk file can be accomplished using the Power Spectrum block (found in DSP Functions group) as follows:

Of course, the resulting averaged waveform may be used by any other block function, not just the displays as shown above. The resulting average could have been written to a disk file, or used along with any other displays or processing functions.

Averaging using the Running Frame Combiner Function

In the Hypersignal RIDE application, as shown in Figure 3, averaging from a disk file may also be accomplished using the Running Frame Combiner block (found in Frame Functions group) as follows:

Note that the contribution of the current frame is much greater using this approach as opposed to the first approach (using Global Frame Mean), and may also have more current frame contribution than the power spectrum method, depending upon the weighting factor used.

Reader Comments | Submit a comment »