arXiv:2609.20331
State-Space-Based FIR Filtering on a Quantum Computer
Roope Salmi, Davide Rocchesso, Vesa Välimäki
Abstract
Many signal processing tasks require intensive computations. Quantum computing promises to accelerate certain tasks, but algorithms must be designed around the limitations of quantum mechanics. This paper provides a quantum implementation of finite impulse response (FIR) filters, which are a widely used tool in classical signal processing. The filter can be parallelized and composed as part of larger quantum algorithms, with potential for speedups using quantum amplitude estimation and future fault-tolerant quantum hardware. To accomplish these properties, we introduce a state-space framework wherein sample-based signal processing is performed with unitary quantum circuits. We present the quantum delay gate as the quantum analog of the delay line in discrete-time signal processing. Unitary circuits intrinsically describe lossless systems, but we also implement lowpass and other bounded filters by emulating a projection operator. The FIR filter implementation is tested and demonstrated with a quantum circuit simulator.
AI Summary
The paper asks how sample-by-sample linear filtering can be implemented coherently on a quantum computer and composed with subsequent processing without intermediate measurement. Signals are amplitude-encoded, while input/output amplitudes and persistent system-state variables occupy orthogonal subspaces. A unitary state-space realization processes each sample, with transpositions routing samples between the signal buffer and processing subspace. The construction implements delays as cyclic permutations and emulates a projection by moving unwanted amplitudes into a delay buffer of length $p$. This reproduces discarding over at most $p$ processing steps, using resources logarithmic in $p$ for the delay. Bounded real FIR filters, satisfying $|H(e^{i\omega})|\leq1$, are implemented through a classical lossless lattice decomposition into rotations and unit delays. An order-$M$ filter uses $M+1$ rotation parameters, obtained from the desired response and a power-complementary filter through spectral factorization and recursive reduction. Additional registers allow the same circuit to process multiple amplitude-encoded signals in superposition. A Qiskit simulation filters two $N=32$ signals with an order-$12$ lowpass filter, using projection delay $p=32$. The transpiled circuit contains 59,292 gates and has depth 38,387; signed output amplitudes are estimated with 1,024-shot Hadamard tests. No quantum amplitude estimation is demonstrated. The paper notes that arbitrary input preparation costs $\mathcal{O}(N)$ gates, circuit length grows with the number of processed samples, and these depths require fault-tolerant hardware. The presented filter uses real coefficients and requires a power complement. Speedups from parallel processing and amplitude estimation are prospective rather than established end-to-end.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- Submitted to ICASSP 2027