Thesis: Noise reduction and acoustic echo cancellation

Background
Handsfree telephony has become a standard infotainment feature in premium vehicles worldwide. The handsfree system allows the driver to safely and conveniently make phone calls while driving. In addition, new in-vehicle speech recognition systems have been introduced in recent years to enable the driver to interact with the vehicle using speech.
Both of these systems (handsfree telephony, and speech) rely on the quality of the speech signals that are captured by the microphone(s) incorporated into the vehicle cabin.
Automakers and AI giants (Google, Amazon) have made large investments in introducing speech systems and voice assistants in vehicles and this branch is currently growing very rapidly.
Problem
Noise in the captured speech signal: The speech signal captured by the microphones is often contaminated by the noise from engines, mechanical vibrations from windows, road, and air-conditioning fan. Trucks are specifically challenging environments due to the large engines, size of the cabin, reverberation, and road-related noise at high driving speeds.
Echo in phone conversation: A telephone conversation is a full duplex telecommunication that consists of a near-end and a far-end component. The sound components from the far-end played on the speaker travel through the cabin and leak to the microphone where they are added with the speech components coming from the driver (near end). This causes the far-end to hear an echo of their voice which is unpleasant and disturbing.
Solution
Noise reduction in the captured speech signal: There are several solutions based on filter theory that have emerged in the last several decades to effectively separate the noise from the desired speech. Recently, classifications based on deep neural networks have emerged 2 (3) too. However, in the case of trucks, as the noise frequency band is quiet low and relatively separated from the speech frequency band, conventional finite impulse response (FIR) filtering has proved successful provided that the details of the filter are correctly designed. Speech-innoise recordings will be provided by Volvo.
Acoustic echo cancelation (AEC): An AEC algorithm aims to reproduce the acoustic path between the speaker (where the far-end speech signal is played) and the microphone and effectively mimic and subtract the echoed components that arrive at the microphone. The acoustic path is usually estimated using machine learning algorithms based on normalized least mean square (NLMS) adaptive FRI filters.
Goal of the thesis
The goal of the thesis is to design an optimized FIR filter to effectively reduce the noise in the captured audio signal. Furthermore, to design machine learning algorithms based on LMS adaptive FIR filters to implement an effective AEC solution for trucks.
The implementation could be done in MATLAB. The results will be published in the thesis and potentially in scientific conferences.
The thesis is meant to produce useful information for Volvo technology AB helping solve one of the most prioritized technical problems while preparing the student for a career in electronic engineering, signal processing, and machine learning.
The infotainment is a very fast growing section of the automotive industry with bright career prospects.
Desirable expertise
  • Signal and systems
  • Machine learning and filter theory
  • Audio/speech signal processing
  • MATLAB
Kick-off date
The thesis can start when there is a successful candidate(s).
Additional info
The scope can be adapted to 1 or 2 students.
The recorded audio files will be provided to the successful candidate by Volvo. If further recordings are necessary, Volvo will facilitate the test setup and tools.
For more information please contact
Alejandro Cortes, GTM, System, Function & Verification, +46 31 3227873

Acerca de nosotros

The Volvo Group is one of the world’s leading manufacturers of trucks, buses, construction equipment and marine and industrial engines under the leading brands Volvo, Renault Trucks, Mack, UD Trucks, Eicher, SDLG, Terex Trucks, Prevost, Nova Bus, UD Bus and Volvo Penta.

Volvo Group Trucks Technology provides Volvo Group Trucks and Business Area's with state-of-the-art research, cutting-edge engineering, product planning and purchasing services, as well as aftermarket product support. With Volvo Group Trucks Technology you will be part of a global and diverse team of highly skilled professionals who work with passion, trust each other and embrace change to stay ahead. We make our customers win.

Queremos conocerte

Proceso de postulación

Postular

Se enviará un correo electrónico de confirmación tan pronto como envíes tu postulación. Después de esto, aún podrás actualizar tu perfil personal.

Entrevista

Si eres seleccionado para una entrevista, te contactaremos para darte mayor información sobre los siguientes pasos del proceso: segunda entrevista, evaluaciones y referencias.

Selección

Todos los candidatos serán notificados cuando se seleccione un candidato final para el puesto de trabajo. Puedes elegir que mantengamos tu currículum en nuestra base de datos y activemos un agente de búsqueda que busque otros trabajos que coincidan con tu perfil.

Inducción

Cuando comiences en tu nuevo puesto de trabajo, recibirás una inducción para ayudarte a formar parte del equipo y comenzar a trabajar con tus tareas de la mejor manera posible.

Testimoniales

Empleos similares