• Hearing protection using sound expo-sure measurement• Hearing test using sma rt phone apps• Automatic adjustment according to the level of hearing.ese features were decided upon through discussions with many mem-bers of the Binaural Hearing Device Study Group, which was formed in October 2019, approximately six months aer the launch of Rion's Tech-nical Development Center. is device is designed based on the concept of pro-viding the “ordinary hearing” we expe-rience for granted using both ears as well as realizing the “incredible hear-ing” by utilizing the sound arriving at our two ears.—It must have been quite the chal-lenge to incorporate so many functions into this device. Can you tell us about how you conquered some of the di culties?Fujisaka: First, the SoC (System on Chip) used in this device is actually a gener-al-purpose SoC used in wireless earbuds and other devices. So the know-how on SoC coding we’d cultivated in our past development eorts on hearing aids didn’t apply. We had to acquire new know-how. Fortunately, we didn't nd it all that hard to implement the acous-tic processing functions for hearing assistance, with the help of a cooper-ating company. In our Binaural Hearing Assist Device that requires true real-time processing, it’s import-ant to minimize processing delays. That proved to be the hard part. e side-branch type signal ow (also used in hearing aids) is known as an eective technique for reducing delays and was part of the device specications. But the signal processing block of a normal wireless earbuds failed to achieve toler-able processing delays, so the block had to be changed. We imposed a constraint, that a stable output of the processed sound had to be ensured at all times without compromise. is modication ate up much of the time set aside for rmware development.Indent the aforementioned MVDR beamformer was computationally ex pen sive and couldn’t be implemented if the calculations were performed in strict accordance with the theory for each processing frame. So we decided to design a beamforming lter in advance, based on the assumption that the target sound was located to the front le or front right. Normally, this lter is posi-tioned before signal processing. But since the ltering process would result in a delay, we devised a method to implement it without aecting the delay by convoluting the ltering process with the gain output from the WDRC system in the frequency domain of the side-branch type signal ow, as shown in Figure 1. As shown in Figure 2, this method improves intelligibility by about 7%, as evaluated by an objective speech intelligibility estimation. It helped us achieve both low delays and ee c tive ness.—How do you think this device, if commercialized, will contribute to society?Fujisaka: Hearing loss has drawn atten-tion in recent years as a risk factor for dementia. is concept model is equipped with functions to monitor noise levels and reduce the loudness of the sound that enters the ear in noisy environments and to reduce impact noise in order to reduce the risk of noise-induced hearing loss. We hope seniors will use this device to enjoy “ordinary hearing” once again and to relieve their social isolation and thereby lowering the risk of dementia and extending healthy life expectancy. is can curb rising cost burdens related to nursing care for growing numbers of people needing such care. I believe it also oers a solution to social issues, such as the decline in productivity in Japan, the pension problem, and the extension of the retirement age.Fig. 1. Signal processing block of the concept modelFig. 2. Results of objective speech intelligibility estimates with and without beamforming processing of sound located to the front right (left figure) and front left (right figure)13
元のページ ../index.html#15