Subband and Frequency-Domain Adaptive Filtering Techniques for Speech Enhancement in Hands-free Communication

The telecommunications sector is characterized by an increasing demand for user-friendliness and interactivity. This explains the growing interest in hands-free communication systems. Signal quality in current hands-free systems is unsatisfactory. To overcome this, advanced signal processing techniques such as the subband and frequency-domain adaptive filter are employed to enhance the signal. These techniques are known to have computationally efficient solutions. Furthermore, thanks to the frequency-dependent processing and adaptivity, highly time-varying systems and signals with a continuously changing spectral content such as speech can be handled. This thesis deals with subband and frequency-domain adaptive filtering techniques for speech enhancement in hands-free communication. The text consists of four parts. In the first part design methods for perfect and nearly perfect reconstruction DFT modulated filter banks are discussed. Part II deals with subband and frequency-domain adaptive filtering. The subband adaptive filter and the PBFDAF-algorithm are discussed. Next, the interrelation between both approaches is studied and a novel subband adaptation scheme is proposed. In part III of the thesis an extension to the PBFDAF algorithm is presented, called the PBFDRAP adaptive filter. The algorithm is analyzed and fast implementation schemes are derived. In the final part we describe applications of our algorithms to the acoustic echo cancellation problem. It is seen that the algorithms discussed in parts I-III can be successfully applied to real-world signal enhancement applications.