Skip to content
Murali RaviMedical Instrumentation

Funded projects · Embedded ML

Voice Biometrics & Edge AI

Biometric security usually assumes a network connection and a server. This project, with Analog Devices, assumes neither: the entire verification pipeline, from feature extraction to decision, runs on an ARM Cortex-A55 inside the vehicle.

Funding
₹6 Lakhs
Projects
1
Span
2025-2026
Case studies
1

Sponsored projects

The project.

Principal InvestigatorActive2025-2026

Design and Implementation of a PoC for Voice Authentication in Cars using an ML approach

A proof-of-concept for in-vehicle speaker verification running entirely at the edge. Produced ECAPA-Lite with a lightweight prosody encoder, quantised to INT8 for deployment on ARM Cortex-A55.

Full case study

Awarded

₹6 Lakhs

Funding agency
Analog Devices, Inc.
Period
2025-2026
Role
Principal Investigator

Featured stories

What the funding produced.

Proof of concept

Edge-Deployable Voice Biometric Authentication

Speaker verification at 2.17% EER, compressed to 1.57 MiB, running on the device itself.

2.17%

Equal Error Rate (float)

2.60%

EER after INT8 quantisation

1.57 MiB

Quantised model size

LIGHTWEIGHT PROSODY ENCODER: A PARALLEL BRANCH FOR SUPRA-SEGMENTAL FEATURESINPUTWaveformBRANCH 01: SPECTRALMFCC → ECAPA-LiteVocal tract shape · segmental detailBRANCH 02: PROSODY · NOVELLightweight EncoderPitchF0 contourEnergyIntensity envelopeDurationTiming & rhythmFUSIONSpeaker embeddingEnhanced discriminationbetween speakersWHY IT ADDS RATHER THAN DUPLICATESProsody varies slowly and encodes speaking style; MFCCs capture vocal tract shape.THE TWO CARRY DIFFERENT IDENTITY INFORMATION: WHICH IS WHY THE BRANCH IMPROVES DISCRIMINATION AT NEGLIGIBLE PARAMETER COST.

Continue

Other domains.

Instrumentation

Medical & Gamma Imaging

Indigenous gamma cameras and intraoperative probes that bring nuclear imaging to the bedside and the operating theatre.

₹274L · 3 projects

FPGA · DSP

Hardware Acceleration

Removing the computational bottleneck that keeps iterative reconstruction out of real-time clinical practice.

Research programme

Industry Programme

Optical Networks & AI/ML

Applying machine learning to optical network design, algorithm development and hardware manufacturing at industrial scale.

₹290L · 2 projects