Demo#

Not Yet

Overview#

Haptic Signal Quality Evaluation is an evaluation framework for AI-generated haptic signals that aims to make “haptic quality” measurable, comparable, and trackable across models, edits, and datasets. Currently building an end-to-end pipeline that ingests generated vibration/haptic waveforms, applies standardized preprocessing, and computes a suite of objective quality metrics that capture both perceptual fidelity (does it feel like the intended effect?) and signal realism (does it resemble real haptic data?). Inspired by how audio research evaluates synthesized sound, the system adapts proven ideas like local alignment-based similarity and global distribution-based scoring into the haptics domain, producing summary scores and diagnostic breakdowns that help researchers understand why one haptic output is better than another. Ultimately, the goal is to provide a repeatable benchmark and reporting format so HapticGen models can be evaluated with the same rigor that audio and vision generation already enjoy.

Why#

As generative models begin producing haptic effects for XR, accessibility, entertainment, and human–computer interaction, the bottleneck shifts from “can we generate something?” to “how do we know it’s good?” Unlike text or images, haptic signals are temporal, hardware-dependent, and highly sensitive to subtle distortions—small changes in amplitude envelopes, frequency content, latency, or transient structure can dramatically change the perceived sensation. Today, many haptic-generation papers still rely on limited subjective demos or ad hoc metrics, making it hard to compare approaches, detect failure modes, or iterate on model improvements. HapticGen Quality addresses this gap by building objective, perceptually motivated metrics and a standardized benchmarking pipeline that can reliably distinguish high-quality signals from artifacts, quantify improvements from model edits, and reveal tradeoffs between fidelity and diversity. By turning haptic evaluation into something reproducible and scalable, the project is meant to accelerate research progress and support real-world deployment where consistent tactile experience matters.