VOCAL’s AEC software has many advantages over open source AEC solutions. One advantage is being able to perform AEC at fullband sampling rates. Open sources solutions may accept fullband sampling rates as input and generate a fullband output, but in between they are downsampling the audio. When the audio is downsampled, information in the upper frequencies is lost.
To their defense, there are practical reasons to perform echo cancellation at lower sampling rates. One reason is the absorption properties of sound. The echo tail length of upper frequencies is much smaller than that of lower frequencies. A second reason is the majority of the signal energy and speech intelligibility is below 8kHz. Finally, the ccomputational complexity of an AEC at 48kHz can be 18 times greater than at 16kHz sampling. So, the conclusion can be, we would be doing a lot of extra work for little gain. We state “can be” because, as discussed in, Efficient Fullband and Wideband AEC Implementations, and Subband Acoustic Echo Cancellation with Variable Echo Tail Lengths, there are solutions that take advantage of the acoustic properties of sound to minimize the computation complexities of fullband AECs. Also, the information above 8kHz can be very meaningful to the audio experience.
VOCAL’s AEC software is a true fullband solution. No information is lost, only quality gained. VOCAL’s software is run-time configurable to the sampling rate of choice.
Figure 1. An AEC result with downsampling
Figure 2. VOCAL’s AEC Result.