Interspeech 2020 2020
DOI: 10.21437/interspeech.2020-2818
|View full text |Cite
|
Sign up to set email alerts
|

Hearing-Impaired Bio-Inspired Cochlear Models for Real-Time Auditory Applications

Abstract: Biophysically realistic models of the cochlea are based on cascaded transmission-line (TL) models which capture longitudinal coupling, cochlear nonlinearities, as well as the human frequency selectivity. However, these models are slow to compute (order of seconds/minutes) while machine-hearing and hearing-aid applications require a real-time solution. Consequently, real-time applications often adopt more basic and less time-consuming descriptions of cochlear processing (gammatone, dual resonance nonlinear) eve… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...

Citation Types

0
1
0

Year Published

2020
2020
2024
2024

Publication Types

Select...
3
2
1

Relationship

2
4

Authors

Journals

citations
Cited by 6 publications
(1 citation statement)
references
References 11 publications
0
1
0
Order By: Relevance
“…The responses of the three ANF types are combined together to yield the final summed AN response 𝑟F, by using weights HNH, MNH and LNH that correspond to the number of HSR, MSR and LSR fibers in a NH periphery (HNH = 13, MNH = 3 and LNH = 3 as reported in Verhulst et al [17]). The CoNNearcochlea, CoNNearIHC and CoNNearANf modules comprise encoder-decoder CNN architectures that can be backpropagated through, thus facilitating the development of individualized audioenhancement methods.2.2 DNN-Based CS-Compensating HA AlgorithmsFrom the NH CoNNear model, we can obtain a HI CoNNear model by retraining the CoNNearcochlea stage via transfer learning to simulate OHC loss[19], and by changing the weights of the different types of ANFs in the CoNNearANf stage to model AN fiber loss, related to CS. The CoNNear HI periphery model can be individualized based on frequency-dependent degrees of OHC loss and CS.…”
mentioning
confidence: 99%
“…The responses of the three ANF types are combined together to yield the final summed AN response 𝑟F, by using weights HNH, MNH and LNH that correspond to the number of HSR, MSR and LSR fibers in a NH periphery (HNH = 13, MNH = 3 and LNH = 3 as reported in Verhulst et al [17]). The CoNNearcochlea, CoNNearIHC and CoNNearANf modules comprise encoder-decoder CNN architectures that can be backpropagated through, thus facilitating the development of individualized audioenhancement methods.2.2 DNN-Based CS-Compensating HA AlgorithmsFrom the NH CoNNear model, we can obtain a HI CoNNear model by retraining the CoNNearcochlea stage via transfer learning to simulate OHC loss[19], and by changing the weights of the different types of ANFs in the CoNNearANf stage to model AN fiber loss, related to CS. The CoNNear HI periphery model can be individualized based on frequency-dependent degrees of OHC loss and CS.…”
mentioning
confidence: 99%