Handbook on Array Processing and Sensor Networks

CHAPTER 8

Acoustic Array Processing for Speech Enhancement

Markus Buck¹, Eberhard Hänsler², Mohamed Krini¹, Gerhard Schmidt¹, and Tobias Wolff¹

¹ Harman/Becker Automotive Systems, Ulm, Germany

² Technische Universität Darmstadt, Darmstadt, Germany

8.1 INTRODUCTION

Today, hands-free functionality has become a standard for acoustic front ends of telephone and speech dialog systems. However, these systems are often applied in adverse acoustic environments where ambient noise as well as acoustic couplings of loudspeaker signals superpose the desired speech signal. Furthermore, the level of the desired speech signal is reduced due to the relatively large distance between speaker and microphones. Therefore, the quality of the microphone signals is poor. Methods for controlling noise and echo without degrading speech quality are still subject of intensive research.

The application of array processing has opened new chances in speech and audio signal processing. There are at least two favorable features: reduction of processing power and/or considerable improvement of system performance.

Speech enhancement procedures demand high amounts of computing power even if they work with a single input channel. This attribute stems from the properties of the electroacoustic environment. Splitting the signal into subbands and, for example, adapting cancellation filters for the subband signals reduces ...