Matrix-product-state compression of quantum-encoded vocal-tract transfer functions.

Vocal-tract transfer functions are densely sampled acoustic representations, making direct quantum-state preparation costly. In this paper, we investigate whether their resonant structure permits compact quantum-compatible encoding. Transfer functions from the Dresden Vocal Tract Dataset are represented as 14-qubit amplitude states and compressed using matrix product states. The resulting states exhibit low cross-partition complexity, allowing a bond-dimension-eight representation to reduce storage by 13.7 times while retaining reconstruction accuracy close to the uncompressed finite-element reference. A complex encoding also preserves phase. These findings show how acoustic structure can reduce the memory required for sequential preparation of quantum-encoded speech representations.