Enhancing Law-Enforcement Audio Transcription: A LoRA-Based Adaptation of Whisper for BWC Footage

2026-07-27

AI Summary

This work addresses the challenge law enforcement agencies face in leveraging body-worn camera (BWC) audio for accountability and systemic review due to the high cost of manual transcription. To overcome this, the authors propose an efficient, domain-adapted speech transcription and structured analysis pipeline tailored for policing contexts. By applying parameter-efficient fine-tuning of the Whisper model via LoRA, combined with 8-bit quantization and gradient checkpointing, the system achieves high-performance deployment on a consumer-grade GPU with only 4 GB of VRAM. A domain ontology–driven symbolic reasoning module further enables automatic generation of evidence-linked event graphs. Experimental results demonstrate significant improvements in transcription accuracy under challenging acoustic conditions—such as sirens and radio interference—and achieve a 93.7% terminology mapping rate, effectively supporting downstream structured analysis.

Abstract

Modern policing faces a "visibility paradox" where law enforcement agencies possess petabytes of Body-Worn Camera (BWC) footage that remains largely unutilized for accountability or systemic review due to the prohibitive labor costs of manual transcription. This research presents a framework for adapting the OpenAI Whisper architecture to the unique acoustic and linguistic challenges of the policing environment. By employing Parameter-Efficient Fine-Tuning (PEFT) through Low-Rank Adaptation (LoRA), we address the significant performance degradation observed in zero-shot models when confronted with high-stress scenarios, sirens, and radio interference. Crucially, we demonstrate that this adaptation is feasible on consumer-grade hardware (Acer Nitro local machine with NVIDIA 4GB GTX GPU) using 8-bit quantization and gradient checkpointing. We further integrate these transcriptions into a symbolic reasoning pipeline using a domain-specific ontology to transform raw audio into evidence-linked incident graphs, achieving a 93.7% lexicon mapping rate for the advancement of procedural justice and transparency.

Problem

• Body-Worn Camera
• audio transcription
• law enforcement
• acoustic challenges
• procedural justice

Innovation

• LoRA
• Whisper
• Parameter-Efficient Fine-Tuning
• Body-Worn Camera
• Ontology-based Reasoning