COCHLEAR IMPLANT AND NORMAL-HEARING LISTENERS

Get Complete Project Material File(s) Now! »

GENERIC ACOUSTIC MODEL

The typical processing steps in an acoustic model are illustrated in Figure 2.1. The first block of the acoustic model (I) mimics the signal processing of CIs, such as the signal bandwidth and analysis filter parameters. It also models aspects of speech processing, such as parameters of envelope extraction and adjustment of signal envelopes (as used in Advanced Combination Encoder [ACE], for example) to some extent. These aspects are combined into one block, as the CI combines them in the speech processor. This block typically consists of filtering the incoming signal into a number of contiguous frequency channels using suitable band-pass filters (BPF). The filtered signal in each channel is no half- or full-wave rectified and then low-pass filtered (usually at 200 – 400 Hz) to establish an envelope for each channel. These processing steps are generalisations of the signal processing used in CIs. The second block (II) is concerned with the generation of synthesis or carrier) signals. The synthesis signal most commonly used is filtered noise bands (Shannon et al., 1995; Dorman et al., 1997a). Synthesis signals embody the assumptions about perception of electrical stimulation. For example, the centre frequencies for filtered noise bands can be chosen according to the electrode positions to model electrode spacing and insertion depth effects (Li and Fu, 2010; Li and Fu, 2007; Baskent and Shannon, 2003; Baskent and Shannon, 2007; Faulkner et al., 2003; Dorman et al., 1997a). The width of the filtered noise bands may model spread of excitation (Bingabr et al., 2008; Blamey, Dowell, Tong and Clark, 1984b), although this aspect is not generally recognised by modellers. The signal envelopes in the different channels (outputs of step I) are used to modulate the noise bands (outputs of step II). Alternatively the signal envelopes modulate white noise, after which filters are applied to the amplitude modulated noise. The modulated noise outputs are combined to arrive at the final signal, which is saved as a sound file, and is presented to normal-hearing listeners. The outputs of each processing step are shown in Figure 2.2.
In order to understand the functional environment of acoustic models, an investigation of clinical and design parameters in current CI systems is needed. A brief overview of the anatomy of the ear and electrophysiology of electrical and acoustical stimulation is also made in order to understand differences which exist on the electrophysiological level, and which may be incorporated in more advanced acoustic models. A comparison of the psychoacoustics of electrical stimulation and acoustic stimulation is needed for a proper construction of synthesis signals. This embodies the assumptions about the perception of sound elicited by electrical stimulation. The choice of synthesis signal is also influenced b theories of hearing, for example the coding of pitch and loudness perceived by the ear. Each one of these aspects will be studied in paragraphs 2.3 to 2.6.

CHAPTER 1 RESEARCH PROBLEM
1.1 INTRODUCTION
1.2 PROBLEM STATEMENT
1.3 RESEARCH QUESTIONS
1.4 APPROACH
1.5 RESEARCH OBJECTIVES
1.6 OVERVIEW OF THESIS
CHAPTER 2 PARAMETERS THAT INFLUENCE PERCEPTION IN COCHLEAR IMPLANT AND NORMAL-HEARING LISTENERS
2.1 INTRODUCTION
2.2 GENERIC ACOUSTIC MODEL
2.3 CI PARAMETERS
2.3.1 Signal processing
2.3.2 Speech processing
2.3.3 Dynamic range compression
2.3.4 Insertion depth and frequency compression effects
2.3.5 Number and spacing of electrodes
2.3.6 Mode of stimulation
2.3.7 Rate of stimulation
2.4 PSYCHOACOUSTICS
2.4.1 Pitch
2.4.2 Intensity, loudness and dynamic range
2.5 ANATOMY
2.5.1 Overview
2.5.2 Speech intelligibility of normal-hearing listener
2.5.3 Acoustic model considerations
2.6 ELECTROPHYSIOLOGY .
2.6.1 Speech intelligibility of normal-hearing listeners
2.6.2 Acoustic model considerations
2.7 CONCLUSION
CHAPTER 3 FRAMEWORK FOR AN ACOUSTIC MODE
3.1 INTRODUCTION .
3.2 FRAMEWORK FOR ACOUSTIC MODELS
3.3 SYSTEM LAYERS
3.3.1 Signal- and speech-processing aspects
3.3.2 Physical implant aspects
3.3.3 Electrical aspects
3.3.4 Electrophysiological aspects.
3.3.5 Perceptual aspects
3.4 SIGNAL PROCESSING .
3.4.1 Block 1: Band-pass filter
3.4.2 Block 2: Extract envelope
3.4.3 Block 3: Compression
3.4.4 Block 4: Current spread
3.4.5 Block 5: Scaling of intensity
3.4.6 Block 6: Loudness growth function
3.4.7 Block 7: Synthesis signals
3.4.8 Modulation of synthesis signal
3.5 POWER SPECTRAL DENSITIES (PSDS) OF PROCESSED SIGNALS .
3.6 CONCLUSION
CHAPTER 4 MODELLING THE ELECTRICAL INTERFACE: EFFECTS OF ELECTRICAL FIELD INTERACTION
4.1 INTRODUCTION
4.2 METHODS
4.2.1 Acoustic models
4.2.2 Experimental methods
4.3 RESULTS
4.3.1 Sentence intelligibility
4.3.2 Consonant intelligibility .
4.3.3 Vowel intelligibility
4.3.4 Effect of modelled current decay
4.3.5 Effect of different compression functions
4.4 DISCUSSION
4.4.1 Asymptote in speech intelligibility
4.4.2 Comparison with other acoustic models
4.4.3 Comparison with CI listener results
4.4.4 Effect of modelled current decay.
4.4.5 Effect of the compression function
4.5 CONCLUSION
CHAPTER 5 MODELLING THE ELECTRICAL INTERFACE: EFFECTS OF SIMULTANEOUS STIMULATION AND COMPRESSION FUNCTION
CHAPTER 6 MODELLING THE PERCEPTUAL LAYER: EFFECTS OF DIFFERENT SYNTHESIS SIGNALS
CHAPTER 7 DISCUSSION
CHAPTER 8 CONCLUSION
REFERENCES