On the Design of Automatic Sound Classification Systems for Digital Hearing Aids

Alexandre E, Álvarez-Perez L, Gil-Pita R, Vicen-Bueno R, Cuadra L. On the Design of Automatic Sound Classification Systems for Digital Hearing Aids. In: AES Convention 126. Audio Engineering Society; 2009. Paper 7735. Available from: https://aes.org/publications/elibrary-page/?id=14931

@inproceedings{Alexandre2009_14931,
  author = {Alexandre, Enrique and Álvarez-Perez, Lorena and Gil-Pita, Roberto and Vicen-Bueno, Raúl and Cuadra, Lucas},
  title = {{On the Design of Automatic Sound Classification Systems for Digital Hearing Aids}},
  booktitle = {AES Convention 126},
  note = {Paper 7735},
  year = {2009},
  month = may,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=14931}
}

TY  - CPAPER
TI  - On the Design of Automatic Sound Classification Systems for Digital Hearing Aids
AU  - Alexandre, Enrique
AU  - Álvarez-Perez, Lorena
AU  - Gil-Pita, Roberto
AU  - Vicen-Bueno, Raúl
AU  - Cuadra, Lucas
T2  - AES Convention 126
M1  - Paper 7735
PY  - 2009
DA  - 2009/05/06
UR  - https://aes.org/publications/elibrary-page/?id=14931
PB  - Audio Engineering Society
LA  - en
AB  - The design of digital hearing aids able to carry out advanced functionalities (such as, for instance, classify the acoustic environment and automatically select the best amplification program for the user's comfort) exhibits a great difficulty. Since hearing aids have to work at very low clock frequency in order to minimize power consumption and maximize life battery, the number of available instructions per second is actually very small. This enforces to design efficient algorithms with a reduced number of instructions. In particular, the paper will focus on three extremely related topics: 1) The design of low-complexity features; 2) The use of automatic feature selection algorithms to optimize the performance of the classifier; and 3) The critical analysis of a variety of different classification algorithms, basically based on their complexity and performance, and determining whether o not they are feasible to be implemented.
ER  -