Y. C. Tong

A parametric formulation of the generalized spectral subtraction method

In this paper, two short-time spectral amplitude estimators of the speech signal are derived based on a parametric formulation of the original generalized spectral subtraction method. The objective is to improve the noise suppression performance of the original method while maintaining its computational simplicity. The proposed parametric formulation describes the original method and several of its modifications. Based on the formulation, the speech spectral amplitude estimator is derived and optimized by minimizing the mean-square error (MSE) of the speech spectrum. With a constraint imposed on the parameters inherent in the formulation, a second estimator is also derived and optimized. The two estimators are different from those derived in most modified spectral subtraction methods, which are predominantly nonstatistical. When tested under stationary white Gaussian noise and semistationary Jeep noise, they showed improved noise suppression results.

Psychophysical studies evaluating the feasibility of a speech processing strategy for a multiple-channel cochlear implant

This paper reports further psychophysical studies on a multiple-channel cochlear implant patient evaluating the feasibility of a speech processing strategy which converts the acoustic fundamental frequency to electric repetition rate, the second-formant frequency to electrode position, and the acoustic amplitude to current level. The first four studies evaluated the use of a special pulse pattern to minimize the loudness variation with electric repetition rate. The chosen pulse pattern consisted of multiple pulses occurring in the first half of each repetitive period (MPP) in contrast to the more conventional pattern with a single pulse per period (SPP). The results showed that MPP approximately equalized the loudness variation with repetition rate. The dynamic range of current, the pitch variation with repetition rate, and the difference limens for repetition rate were found to be similar to MPP and SPP. Two other studies investigated interaction between electrode position and repetition rate (RR). The first of these showed that the patient could make use of information provided by rising or falling RR trajectories superimposed on individual electrodes or electrode trajectories as an indicator of the direction of intonation variation. The second of these studies showed that the dissimilarities amongst the hearing sensations produced by steady-state stimuli differing in electrode position and repetition rate were characterized by two perceptual components, relating to the two electric parameters, respectively.

A multiple electrode cochlear implant

The University of Melbourne Departments of Otolaryngology and Electrical Engineering (UMDOLEE) receiving and stimulating component of a multiple-electrode cochlear implant hearing prosthesis produces constant stimulation. It has a stimulating pulse shape that minimizes the production of toxic substances and loss of metal from the electrodes, and this is achieved with a biphasic rectangular waveform where the first phase is negative with respect to ground. The duration of each stimulus phase in 180 msec, which is long enough to allow low levels of current stimulation, and short enough to permit rates of 1000 pulses/second to be achieved. In order to be consistent with our present understanding of the perception of pitch, the device permits the independent stimulation of a number of electrodes. Furthermore, to electrically isolate the stimulus to small areas, there is the capacity to vary the current and set the threshold independently at individual electrodes. The phase and amplitude of the pulses to neighbouring electrodes with also be varied to assist in localizing the current flow. The pattern to stimulation to individual or groups of electrodes can also be altered to enable studies to be carried out to determine ways of conveying frequency and intensity information over a more normal dynamic range.

Psychophysical studies for two multiple-channel cochlear implant patients.

Psychophysical studies were conducted on two multiple-channel cochlear implant patients to examine the nature of the hearing sensations produced by electrical stimulation of auditory nerve fibers using electrodes at different sites in the scala tympani (one electrode at a time). Both time-invariant stimuli, whose parameter values did not vary in time, and time-varying stimuli, specified by a linear variation in parameter values, were used. A sharpness ranking study using time-invariant signals suggested that the hearing sensations produced by different electrodes varied from dull to sharp in an apical to basal direction in the scala tympani. A categorization study showed that the hearing sensations produced by two adjacent electrodes (1.5 mm apart) were rarely confused for a restricted range of time-invariant pulse rates. Discriminability studies by a same-different procedure for stimuli with pulse rate below 250 pps showed: (1) relative difference limens of 6% to 12% for time-invariant pulse rates, and 9% and 13% for time-varying pulse rates; (2) stimuli with time-varying electrode position differing in the direction of electrode trajectory were readily discriminated; and (3) the discrimination of time-varying pulse rates deteriorated with decreases in the duration of the variation, while the discriminability of single-electrode trajectories was the same for the three durations: 25, 50, and 100 ms. A speech processing strategy was also proposed on the bases of these results.

Analysis and design of power efficient class D amplifier output stages

A Class D amplifier comprises a pulse width modulator and an output stage. In this paper we analyze the power dissipation mechanisms and derive the overall power efficiency of the output stage realized using the finger and waffle layouts. We compare the relative merits of these layouts; we propose two design methodologies to determine the aspect ratios of the transistors in the output stage for optimum power efficiency (optimum for a given fabrication process, supply voltage and load resistance): (1) optimization to a single modulation index point and (2) optimization to a range of modulation indexes. For the design of an output stage with optimum power efficiency (and small IC area), we recommend optimization to a range of modulation indexes and a layout realized by the waffle structure. The theoretical analysis and derivations are verified on the basis of computer simulations and measurements on fabricated prototype ICs.

Speech processing method and apparatus

Signal processing system for converting a speech signal into a data signal for controlling a hearing prosthesis having an implanted electrode array adapted to stimulate the auditory nerve fibers of a patient by the application of electrical currents to selected electrodes in the array. The system generates an input signal current corresponding to a received speech signal. The amplitude and frequency of the fundamental voicing component of the speech signal is approximated as are the amplitude and frequency of at least one formant of the speech signal. A programmable microprocessor produces instructions which cause the application of electrical currents to selected groups of electrodes in the array with or without delays between the stimulation of each electrode in the groups. The microprocessor is programmable with data defining a predetermined relationship between each group of electrodes and a selected range of at least one formant frequency and with data defining a predetermined relationship between another formant frequency and the delay between stimulation of each electrode in each said group based on psychophysical testing of the patient. Selection of electrodes based on the estimated frequency of the formants produces the desired percepts in the auditory-like sensations generated in the patient. The microprocessor is further programmable to produce data which determines the level of stimulation of each selected group of electrodes and determines the delay between stimulation of electrodes in each group dependent on the estimated amplitude of formants of the speech signal as well as on predetermined data relating to the sensitivity of each electrode implanted in the patient.

Forward masking patterns produced by intracochlear electrical stimulation of one and two electrode pairs in the human cochlea

Three psychophysical forward masking studies were conducted on a multichannel cochlear implant patient. The first study investigated the masking pattern produced by a bipolar electrode pair at different stimulus currents. It was found that the masking pattern for a single-masker bipolar electrode pair had a maximum located at an electrode position where the masker and probe coincided. The spread of the masking pattern was not symmetrical about the maximum. The amount of masking decreased very rapidly toward the apical direction and less rapidly toward the basal direction from the position of the maximum. As the stimulus current increased, the amount of masking at the maximum increased and the masking pattern broadened toward the base. The second study investigated the masking pattern produced by the activation of single bipolar electrode pairs with different spatial extents. The spatial extent of a bipolar electrode pair is defined as the distance between the apical and basal electrode members of the bipolar pair. With a small spatial extent (1.5 mm), the more basal electrode pairs (higher threshold and smaller dynamic range) produced broader masking patterns than the more apical electrode pairs (lower threshold and wider dynamic range), suggesting that there was more current spread at the basal region. With a larger spatial extent (4.5 mm), an additional secondary masking maximum was observed in the vicinity of the apical electrode member of the masker; this was observed only when the apical electrode member lay within the low-threshold apical region. The third study investigated the masking patterns produced by two loudness balanced bipolar masker electrode pairs activated within a stimulus period (inverse of the pulse repetition rate). The biphasic current pulses delivered to the two electrode pairs were nonoverlapping in time. It was found that, at any probe electrode position, the amount of masking produced by the two combined bipolar electrode pairs approximately followed the greater of the two maskings produced respectively by the two individual bipolar masker electrode pairs.

A preliminary report on a multiple-channel cochlear implant operation.

This is a publisher’s version of an article published in The Journal of Laryngology & Otology

Dimensions of the scala tympani in the human and cat with reference to cochlear implants

The width, height, and cross-sectional area of the scala tympani in both the human and cat were measured to provide dimensional information relevant to the design of scala tympani electrode arrays. Both the height and width of the human scala tympani decrease rapidly within the first 1.5 mm from the round window. Thereafter, they exhibit a gradual reduction in their dimension with increasing distance from the round window. The cross-sectional area of the human scala tympani reflects the changes observed in both the height and width. In contrast, the cat scala tympani exhibits a rapid decrease in its dimensions over the first 6 to 8 mm from the round window. However, beyond this point the cat scala tympani also exhibits a more gradual decrease in its dimensions. Finally, the width of the scala tympani, in both human and cat, is consistently greater than the height.

The perception of temporal modulations by cochlear implant patients

The perception of temporal modulations of pulsatile electric stimuli was measured in seven cochlear implant patients using the Cochlear Pty. Limited prosthesis. Four patients were postlingually deafened adults and three patients were young adults who were deafened very early in life. The first study measured detection thresholds for modulated pulse duration for a series of modulation frequencies and pulse rates. The shape of the detection thresholds as a function of modulation frequency, the temporal modulation transfer function, often resembled a low-pass filter with a 50-100-Hz cut-off frequency. Thresholds did not markedly vary across the different pulse rates for most patients. Thresholds were less than 10%-20% of the range of usable hearing for most patients. The second study compared detection thresholds for modulated pulse durations around different reference pulse durations: 50, 100, and 300 microseconds. Detection thresholds were generally proportional to the different reference pulse durations. The third study measured difference limens for the discrimination of modulation depth. The difference limens were similar to the detection thresholds for the same reference pulse duration and pulse rate. The three patients deafened very early in life showed more within-group variation in performance, and their overall levels of performance were poorer than those of the postlingually deafened adults.