A warped filter implementation for the loudness enhancement of speech

Cellular phones and small hand-held audio devices have limited power configuration with high current drain audio speaker capabilities. For manufacturing purposes, and cost savings, the audio speakers chosen usually balance product requirements to manufacturing costs. Larger less expensive speakers are usually integrated into the product since smaller high quality power efficient speakers are more expensive. Much of the current focus in industry technology has been better speaker design, or more efficient power amplifiers to minimize battery drain for speaker phone operations. No energy conservation schemes directly operate on the speech signal. The question we address in this dissertation is how to make speech sound louder without increasing the signal energy.; We propose a real-time warped filter which exploits the psychoacoustic nature of the auditory system to enhance the perception of loudness without adding energy. The frequency resolution of sound in the human auditory system is on a non-linear scale called the critical band scale. The critical band concept in auditory theory states that for a constant energy bandwidth product, loudness increases when a critical band is exceeded. A warped filter is proposed and developed to elevate the perception of loudness by applying nonlinear bandwidth expansion to the formant regions of vowels in accordance with the critical band scale. This is the first known study to propose an algorithm which elevates the perception of loudness without adding energy. It is also the first known study to define a filter which adjusts formant bandwidths on a critical band scale, and to use a warped filter for speech enhancement. The underlying technique is an extension of the linear bandwidth broadening technique used for speech modelling in speech recognition, perceptual noise weighting, and vocoder post-filter designs. It is a pole-displacement model, which is a computationally efficient technique, and is included in the linear transformation of the warped filter coefficients. In a warped recursive filter, a coefficient or filter transformation is necessary to avoid un-realizable time dependencies. In this thesis we include the pole displacement model in a warped filter implementation for formant critical bandwidth expansion. The inclusion of a warped pole displacement model for nonlinear bandwidth expansion in the filter was motivated from the critical band concept of hearing. The filter implementation has been inspired by the biological representation of loudness in the peripheral auditory system, and subjective listening tests confer that a noticeable improvement up to 2dB is attainable.