Advanced Speech Encoding (ASE)

Program Manager: Dr. Mari Maeda

Approach:

The past 50 years have seen great strides in voice encoder (vocoder) development, but ultra-low bit rate (ULBR) voice encoding at 300 bits/second (bps) remains a significant challenge. Specifically, ULBR vocoders to date have not delivered adequate aural quality, intelligibility, and speaker recognizability, and these shortcomings are magnified in acoustically harsh (e.g., noisy, reverberant) environments.

The approach taken by the Advanced Speech Encoding (ASE) Program is to use novel noise-immune sensors as an adjunct to the acoustic signal and signal processing (Figure 1). Such novel sensors and other sensors will also be explored with respect to their potential to enable subvocalic/subauditory (nonacoustic) modes of speech and as an alternative means of communication in acoustically harsh and in dangerous military environments where covertness is required.

Figure 1. Sensing both the speech audio signal and glottal motion enables ultra-low bit rate speech encoding.