Processor Mapping Research Articles

Next day cochlear implant switch-on: the New Zealand experience

ABSTRACT The objective of this study was to review the practice of next-day cochlear implant activation, or switch-on, after surgery for adult cochlear recipients in northern New Zealand. A retrospective observational study compared next-day and three-week post-surgical outcomes related to early complications, electrode impedance values, speech perception outcomes and the number of patient appointments and speech processor MAPs to the point of stabilisation. Over a five-year period, 167 consecutively implanted recipients were included in this review. Results indicate that less time from surgery to switch-on had no impact on the incidence of wound or fitting complications, and resulted in lower initial impedance measures, fewer hours of follow-up visits and fewer MAPs during the subsequent nine months. Speech perception outcomes did not differ between the early and later switch-on groups. General satisfaction with next-day switch-on practices has resulted in consistent use of this protocol with high patient and clinician satisfaction and improved clinical efficiencies.

Energy-Aware Scenario-Based Mapping of Deep Learning Applications Onto Heterogeneous Processors Under Real-Time Constraints

To cope with the increasing demand for deep learning applications in embedded systems, emerging embedded devices tend to equip multiple heterogeneous processors, including GPU and deep learning hardware accelerator, called neural processing unit (NPU). It becomes popular to run multiple deep learning (DL) applications simultaneously to provide several functionalities. In this work, we assume that applications have real-time constraints that may vary at run time. While extensive studies have been conducted recently to find an efficient mapping of multiple DL applications on various hardware platforms, they do not consider the constraints imposed by the NPU and the associated software development kit (SDK) in a real embedded platform. In this paper, we propose a novel energy-aware mapping methodology of multiple DL applications onto a real embedded system that has multiple heterogeneous processors. The objective is to minimize energy consumption while satisfying the real-time constraints of all applications. In the proposed scheme, we first select Pareto-optimal mapping solutions for each application. Then mapping combination is explored, considering the scenario that indicates the dynamism of applications while satisfying the constraints. Also, we reduce energy consumption by tuning the frequency of processors. We could satisfy up to <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$40\%$</tex-math></inline-formula> higher deadline constraints and reduce the energy consumption by <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$22\%$</tex-math></inline-formula> <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\sim$</tex-math></inline-formula> 31% compared to the static mapping methods with real-life applications and different scenarios on a real platform.

Hybrid MPI-communication for the multi-angular SN parallel sweep on 3-D regular grids

The Koch-Baker-Alcouffe (KBA) algorithm has been widely used in solving the Boltzmann transport equation in parallel. To improve the scaling efficiency, the sweep is executed from the four corners of a 2-D processor map. Collisions take place when a processor has multiple tasks ready to be solved at one step. In order to handle these collisions, the 2-D processor map is divided into several zones that are defined as sub-processor maps with different octant sweep orders. In order to ensure the communications of these collided processors safe and efficient, four kinds of hybrid MPI-communication algorithms have been analyzed. Hybrid MPI-communications with processor slice achieve better scaling than the ones with block slice. The buffered-blocking communication with processor slice (BBP) achieves the best scaling efficiency. The comparisons of weak-scaling parallel efficiency between the standard KBA algorithm and BBP algorithm for multi-angular SN sweep have been made. The tests are performed on Tianhe-2 super computer by using 102–104 processors. For the test using 104 processors, the weak scaling efficiency is improved by over 13 percent compared to the standard KBA algorithm. The BBP and KBA algorithms have been applied to perform the pin-by-pin calculation of 3-D PWR core, which indicates that the BBP algorithm has the same accuracy with the standard KBA algorithm but better efficiency.

Effects of Site-Specific Level Adjustments on Speech Recognition With Cochlear Implants

Modulation detection thresholds (MDTs) vary across stimulation sites in a cochlear implant (CI) electrode array in a manner that is subject and ear specific. Previous studies have demonstrated that speech recognition with a CI can be improved by site-selection strategies, where selected stimulation sites with poor modulation sensitivity are removed from a subject's processor MAP. Limitations of site-selection strategies are that they can compromise spectral resolution and distort frequency-place mapping because the frequencies assigned to the removed sites are usually reallocated to other sites, and site bandwidths are broadened. The objective of the present study was to test an alternative approach for rehabilitation that aimed at improving the across-site mean MDTs by adjusting stimulation parameters at the poorly performing sites. On the basis of previous findings that modulation detection contributes to speech recognition and improves significantly with stimulus level, the authors hypothesized that modulation sensitivity at the poor sites could be improved by artificially increasing stimulation levels at those sites in the speech processor, which then would lead to improved speech recognition. Nine postlingually deafened ears implanted with Nucleus CIs were evaluated for MDTs, absolute-detection threshold levels (T levels), and the maximum loudness levels (C levels) on each of the available stimulation sites. For each ear, the minimum stimulation level settings in the speech-processor MAP were raised by 5%, and alternatively by 10%, of the dynamic range (DR) from true thresholds on five stimulation sites with the poorest MDTs. For comparison, a 5% level raise was applied globally to all stimulation sites. The C levels were fixed during these level manipulations. MDTs at the five poorest stimulation sites were compared at 20% DR before and after the level adjustments. Speech-reception thresholds (SRTs), that is, signal to noise ratios required for 50% correct speech recognition, were evaluated for these MAPs using CUNY sentences. The site-specific level-adjusted MAPs were compared with the global-level-adjusted MAP and the MAP without level adjustment. The effects on speech recognition of adjusting the minimal stimulation level settings on the five poorest stimulation sites were also compared with effects of removing these sites from the speech-processor MAP. The 5% level increase on the five electrodes with the worst MDTs resulted in an improvement in the group mean SRT of 2.36 dB SNR relative to the MAP without level adjustment. The magnitude of level increase that resulted in the greatest SRT improvement for individuals varied across ears. MDTs measured at 20% DR significantly improved on the poor sites after the level adjustment that resulted in the best SRT for that ear was applied. Increasing the minimal stimulation levels on all stimulation sites or removing sites selected for rehabilitation, the parsimonious approaches, did not improve SRTs. The site-specific adjustments of the T level settings improved modulation sensitivity at low levels and significantly improved subjects' SRTs. Thus, this site-rehabilitation strategy was an effective alternative to site-selection strategies for improving speech recognition in CI users.

Using Temporal Modulation Sensitivity to Select Stimulation Sites for Processor MAPs in Cochlear Implant Listeners

Previous studies in our laboratory showed that temporal acuity as assessed by modulation detection thresholds (MDTs) varied across activation sites and that this site-to-site variability was subject specific. Using two 10-channel MAPs, the previous experiments showed that processor MAPs that had better across-site mean (ASM) MDTs yielded better speech recognition than MAPs with poorer ASM MDTs tested in the same subject. The current study extends our earlier work on developing more optimal-fitting strategies to test the feasibility of using a site-selection approach in the clinical domain. This study examined the hypothesis that revising the clinical speech processor MAP for cochlear implant (CI) recipients by turning off selected sites that have poorer temporal acuity and reallocating frequencies to the remaining electrodes would lead to improved speech recognition. Twelve CI recipients participated in the experiments. We found that site selection procedure based on MDTs in the presence of a masker resulted in improved performance on consonant recognition and recognition of sentences in noise. In contrast, vowel recognition was poorer with the experimental MAP than with the clinical MAP, possibly due to reduced spectral resolution when sites were removed from the experimental MAP. Overall, these results suggest a promising path for improving recipient outcomes using personalized processor-fitting strategies based on a psychophysical measure of temporal acuity.

Psychophysically based site selection coupled with dichotic stimulation improves speech recognition in noise with bilateral cochlear implants

The ability to perceive important features of electrical stimulation varies across stimulation sites within a multichannel implant. The aim of this study was to optimize speech processor MAPs for bilateral implant users by identifying and removing sites with poor psychophysical performance. The psychophysical assessment involved amplitude-modulation detection with and without a masker, and a channel interaction measure quantified as the elevation in modulation detection thresholds in the presence of the masker. Three experimental MAPs were created on an individual-subject basis using data from one of the three psychophysical measures. These experimental MAPs improved the mean psychophysical acuity across the electrode array and provided additional advantages such as increasing spatial separations between electrodes and/or preserving frequency resolution. All 8 subjects showed improved speech recognition in noise with one or more experimental MAPs over their everyday-use clinical MAP. For most subjects, phoneme and sentence recognition in noise were significantly improved by a dichotic experimental MAP that provided better mean psychophysical acuity, a balanced distribution of selected stimulation sites, and preserved frequency resolution. The site-selection strategies serve as useful tools for evaluating the importance of psychophysical acuities needed for good speech recognition in implant users.

Across-site patterns of modulation detection: relation to speech recognition.

The aim of this study was to identify across-site patterns of modulation detection thresholds (MDTs) in subjects with cochlear implants and to determine if removal of sites with the poorest MDTs from speech processor programs would result in improved speech recognition. Five hundred millisecond trains of symmetric-biphasic pulses were modulated sinusoidally at 10 Hz and presented at a rate of 900 pps using monopolar stimulation. Subjects were asked to discriminate a modulated pulse train from an unmodulated pulse train for all electrodes in quiet and in the presence of an interleaved unmodulated masker presented on the adjacent site. Across-site patterns of masked MDTs were then used to construct two 10-channel MAPs such that one MAP consisted of sites with the best masked MDTs and the other MAP consisted of sites with the worst masked MDTs. Subjects' speech recognition skills were compared when they used these two different MAPs. Results showed that MDTs were variable across sites and were elevated in the presence of a masker by various amounts across sites. Better speech recognition was observed when the processor MAP consisted of sites with best masked MDTs, suggesting that temporal modulation sensitivity has important contributions to speech recognition with a cochlear implant.

Effects of electrode configuration on cochlear implant modulation detection thresholds

Cochlear implant function, as assessed by psychophysical measures, varies from one stimulation site to another within a patient's cochlea. This suggests that patient performance might be improved by selection of the best-functioning sites for the processor map. In evaluating stimulation sites for such a strategy, electrode configuration is an important variable. Variation across stimulation sites in loudness-related measures (detection thresholds and maximum comfortable loudness levels), is much larger for stimulation with bipolar electrode configurations than with monopolar configurations. The current study found that, in contrast to the loudness-related measures, magnitudes of across-site means and the across-site variances of modulation detection thresholds were not dependent on electrode configuration, suggesting that the mechanisms underlying variation in these various psychophysical measures are not all the same. The data presented here suggest that bipolar and monopolar electrode configurations are equally effective in identifying good and poor stimulation sites for modulation detection but that the across-site patterns of modulation detection thresholds are not the same for the two configurations. Therefore, it is recommended to test all stimulation sites using the patient's clinically assigned electrode configuration when performing psychophysical evaluation of a patient's modulation detection acuity to select sites for the processor map.

Implementation of MAC-based RTL module for Inverse DCT in H.264/AVC

In this paper, we implemented the MAC-based RTL module for inverse DCT in H.264/AVC to improve applicability, to reduce processing time and utilize resources. The paper highlights design of FU architecture, its interconnection topology, regular formula of inverse DCT and array processor mapping as well as MAC-based RTL module constructing. Multi-directional FUA and FPGA were presented along with an evaluated performance and simulation result. Hence, the paper encompasses design of single FU that was verified with the performance test at maximum frequency 200 MHz; the designed 4-by-4 FUA operates over 100 MHz. The proposed multi-directional FU can be extended to n-by-n FUA that functionality can be extended to next video coding standard (H.265/HEVC).

Heuristics for Flexible CMP Synthesis

Flexible Chip Multiprocessor (CMP) systems are implemented on field programmable devices to exploit both task-level parallelism and architecture customization for parallel programs. The idea is to simultaneously allocate processor resources, map and schedule tasks to them, and to allocate one or several intertask communication resources such that the throughput or execution time is optimized. The design space of such systems is huge, requiring means to automatically optimize design parameters so as to facilitate wide and disciplined explorations. The complexity resulting from corresponding system modeling necessitates the use of optimization heuristics to cope with excessively long runtime for large problem instances. This paper provides a formal proof for the existence of optimum linear time synthesis algorithms for one of two classes of problem instances, and proceeds to present three greedy-like heuristics which exploit the structure of the synthesis problem. A comparison of results for real-time and non-real-time parallel programs is given against integer linear programming, where a synthesis strategy is proposed to achieve good results.

Relation of Electrically Evoked Compound Action Potential Thresholds to Behavioral T- and C-Levels in Children with Cochlear Implants

This study examined the relation of electrically evoked compound action potential thresholds obtained using neural response telemetry (NRT) to T- and C-levels in children's speech processor programs optimized for recognition of very soft to loud sounds while ensuring tolerance of very loud sounds. Forty-one children (age 2 to 14 yr) with stable electrical hearing participated. All children were Nucleus 24 System recipients and attended one of three auditory-oral schools that have on-site pediatric audiologists experienced at cochlear implant programming. Speech processor MAPs were created and adjusted over a period of months until aided warble-tone thresholds were between 10 and 30 dB HL at octave frequencies between 250 and 4000 Hz, and understanding of speech was maximized for many listening situations. At least 1 yr postactivation, visual (vNRT) and predicted (tNRT) thresholds were obtained on 9 to 11 electrodes and compared to each child's T- and C-level values on these electrodes in their MAPs. Test-retest stability of NRT thresholds was compared for two test sessions 1 mo apart. NRT-based evoked compound action potential thresholds could be obtained from 36 of the 41 children. vNRT and tNRT test-retest reliability was high; average correlation coefficients (r) across subjects were 0.90 (range: 0.64 to 0.99) and 0.88 (range: 0.31 to 1.00), respectively. Group average correlation coefficients between vNRT and T-level, vNRT and C-level, tNRT and T-level, and tNRT and C-level were low (0.18, 0.21, 0.24, and 0.26, respectively). Group mean tNRT thresholds were four current levels lower than the group mean vNRT thresholds. Subsequent analysis was performed with the vNRT thresholds because the range of test-retest correlation coefficients for individual subjects was narrower than with tNRT. Hierarchical linear modeling was used to determine if vNRT could be used to predict T- and C-levels. This analysis indicated a significant average relation between vNRT and T-levels and between vNRT and C-levels, but significant heterogeneity in the individual-level estimates of those relations. In other words, subjects varied significantly in the size of the relation between their individual vNRT values and both T- and C-levels. Attempts to account for that heterogeneity did not identify any subject characteristics that were significantly related to the individual-level parameters. The position of the group average vNRT and tNRT thresholds in the upper half of the dynamic range between Ts and Cs agrees with previous studies. The fact that the profile of vNRT thresholds did not parallel the profiles of Ts and Cs across electrodes for most children suggests that simply shifting the NRT profile to select T- and C-levels in initial MAPs is likely to result in a loudness imbalance for certain speech frequencies and/or tolerance issues for many children. This was verified by the hierarchical linear modeling analysis, which showed substantial and significant heterogeneity in the relations between vNRT and T-levels and between vNRT and C-levels. In summary, vNRT is not related to T- or C-levels in a simple and uniform way that would allow it to guide MAP fine tuning with any precision. Consequently, it is recommended that MAP fine tuning be based on the child's behavioral responses on individual electrodes.

Electric Charge Requirements of Pediatric Cochlear Implant Recipients Enrolled in the Childhood Development After Cochlear Implantation Study

To evaluate mapping characteristics of children with cochlear implants who are enrolled in the Childhood Development After Cochlear Implantation (CDACI) multicenter study. Longitudinal evaluation during 24 months of speech processor maps of children with cochlear implants prospectively enrolled in the study. Six tertiary referral centers. One hundred eighty-eight children enrolled in the CDACI study who were 5 years old or younger at the time of enrollment. Of these children, 184 received unilateral implants, and 4 received simultaneous bilateral implants. Children attended regular mapping sessions at their implant clinic as part of the study protocol. Maps were examined for each subject at 4 different time intervals: at device activation and 6, 12, and 24 months postactivation. Mean C/M levels (in charge per phase) were compared for 4 different time intervals, for 3 different devices, for 6 different implant centers, and for children with normal and abnormal cochleae. All 3 types of implant devices demonstrate significant increases in C/M levels between device activation and the 24-month appointment. Significant differences in mean C/M levels were noted between devices. Children with cochlear anomalies demonstrate significantly greater C/M levels than children with normal cochleae. The CDACI study has enabled us to evaluate the mapping characteristics of pediatric patients who use 3 different devices and were implanted at a variety of implant centers. Analysis of such data enables us to better understand the mapping characteristics of children with cochlear implants.

Das Refraktärverhalten des elektrisch stimulierten Hörnervs

Electrically evoked compound action potentials (TECAP) of the auditory nerve can be recorded in cochlear implants. TECAP thresholds are used to predict threshold levels for speech processor maps. The auditory nerve's refractory properties can influence these levels. Recovery functions were investigated at 84 stimulation sites in 14 patients who had Nucleus CI24 implants; neural response telemetry (NRT) and a modified forward-masking technique were used for these investigations, introducing the reference masker-probe interval (MPI). An interval between 300 and 375 micros was found to be suitable as the reference MPI in our study. The median of the absolute refractory period was determined as 390 s and the median time constant of the recovery function, at 425 s. In practice, a reference MPI of 300 s is suggested for measurement of recovery and amplitude growth functions. As up to now the amplitude growth function has been measured at 500 s and thus mostly in a relatively refractory condition, the refractory behaviour should influence the TNRT. In addition, it was possible to explain the shape of standard forward-masking recovery functions with reference to the latency shift of the neural response.

A flexible processor mapping technique toward data localization for block-cyclic data redistribution

Array redistribution is usually needed for more efficiently executing a data-parallel program on distributed memory multicomputers. To minimize the redistribution data transfer cost, processor mapping techniques were proposed to reduce the amount of redistributed data elements. Theses techniques demand that the beginning data elements on a processor not be redistributed in the redistribution. On the other hand, for satisfying practical computation needs, a programmer may require other data elements to be un-redistributed (localized) in the redistribution. In this paper, we propose a flexible processor mapping technique for the Block-Cyclic redistribution to allow the programmer to localize the required data elements in the redistribution. We also present an efficient redistribution method for the redistribution employing our proposed technique. The data transfer cost reduction and system performance improvement for the redistributions with data localization are analyzed and presented in our experimental results.

Psychophysical assessment of stimulation sites in auditory prosthesis electrode arrays

Auditory prostheses use implanted electrode arrays that permit stimulation at many sites along the tonotopic axis of auditory neurons. Psychophysical studies demonstrate that measures of implant function, such as detection and discrimination thresholds, vary considerably across these sites, that the across-site patterns of these measures differ across subjects, and that the likely mechanisms underlying this variability differ across measures. Psychophysical and speech recognition studies suggest that not all stimulation sites contribute equally to perception with the prosthesis and that some sites might have negative effects on perception. Studies that reduce the number of active stimulation sites indicate that most cochlear implant users can effectively utilize a maximum of only about seven sites in their processors. These findings support a strategy for improving implant performance by selecting only the best stimulation sites for the processor map. Another approach is to revise stimulation parameters for ineffective sites in an effort to improve acuity at those sites. In this paper, we discuss data supporting these approaches and some potential pitfalls.

Compiler-directed application mapping for NoC based chip multiprocessors

The problem attacked in this paper is one of automatically mapping an application onto a Network-on-Chip (NoC) based chip multi-processor architecture in a locality-aware fashion. The proposed compiler approach has four major steps: task scheduling, processor mapping, data mapping, and packet routing. Our experimental result clearly shows that the proposed framework reduces energy consumption of our applications significantly (27.41% on average over a pure performance oriented application mapping strategy) as a result of improved locality of data accesses.

Measuring the Refractoriness of the Electrically Stimulated Auditory Nerve

Intracochlear recordings in cochlear implant recipients provide access to the electrically evoked compound action potential (ECAP). ECAP thresholds are potential predictors of speech processor map’s threshold and comfortable loudness levels. The auditory nerve’s refractory properties can influence these levels due to high-rate stimulation with interpulse intervals within the relative refractory period. Recovery functions were investigated at 84 stimulation sites in 14 Nucleus CI24 recipients using neural response telemetry and a modified forward masking technique. This technique introduces a reference masker-probe interval (MPI). In our study, an appropriate value between 300 and 375 µs was determined for this reference MPI, and the use of a reference MPI of 300 µs is suggested for recovery and amplitude growth functions. A median absolute refractory period of about 390 µs and a median time constant of about 425 µs were obtained by fitting an exponential model to the data. Hence, the auditory nerve is usually in relative refractory state when standard neural response telemetry forward masking is selected because of its default MPI of 500 µs. This can bias the measurement of ECAP thresholds. Additionally, the shape of standard forward masking recovery functions was explained by the influence of latency shift of the neural response.

An Efficient Communication Scheduling Method for the Processor Mapping Technique Applied Data Redistribution

Array redistribution is usually required for more efficiently executing a data-parallel program on distributed memory multi-computers. In performing array redistribution using synchronous communication mode, data communications among the processors should be properly arranged to avoid incurring higher data transfer cost. Some efficient communication scheduling methods for the Block-Cyclic redistribution have been proposed. On the other hand, the processor mapping technique can help reduce the data transfer cost of redistribution. To avoid degrading the benefit of data transfer cost reduction, it is needed to construct optimal communication schedules for the redistribution in which the processor mapping technique is applied. In this paper, we present a unified approach to constructing optimal communication schedules for the processor mapping technique applied Block-Cyclic redistribution. The proposed method is founded on the processor mapping technique and can more efficiently construct the required communication schedules than other optimal scheduling methods.

Genetic Algorithms for Adaptive Psychophysical Procedures: Recipient-Directed Design of Speech-Processor MAPs

The goal of the research is to evaluate the application of genetic algorithms (GAs) in listener-directed optimization of audio-processing designs. We hypothesize that cochlear-implant recipients can use a GA-guided adaptive psychophysical search procedure to select useful designs from among a large number of speech processor MAPS. An adaptive psychophysical procedure was developed in which a listener's preferred four out of eight speech processor MAPs were updated according to a genetic algorithm. Experiments involving cochlear-implant recipients were conducted to characterize both the convergence behavior of the adaptive procedure as well as properties of the MAPs optimized by the recipient. Results from five cochlear-implant recipients indicate that the adaptive procedure converges to useful speech processor MAPs within twenty iterations. The results suggest a means whereby a potentially large number of audio-processing designs can be searched efficiently by a human listener without requiring excessive amounts of feedback or prior knowledge about the listener's preferences. In the case of cochlear-implant recipients, it may be possible to use this procedure as an aid to the clinician in the fitting of a speech processor MAP.

Recipient-directed design of speech processor MAPs

A method is introduced whereby implant recipients can search over a large number of speech processor MAPs to identify one(s) they prefer. Using the genetic algorithm (GA), the method searches over multiple MAP parameters simultaneously, rather than over each parameter individually. The method can include as many as 8000 MAPs without significantly increasing search time or degrading search performance. Results are presented for five recipients who used the method to design a speech processor MAP from among 256 options.