NASP NeuroVoice: Development of Brain-Mimicking Chip for Improved Hearing and Speaking Experience for Millions of Users

在Polyn Technology，我们从人脑中获得灵感。我们的眼睛对千兆字节的视觉信息进行前处理，传递给大脑的数据减少到原来的1/10，从而减少了不必要的工作。因此，拥有800亿个神经元的大脑只消耗20瓦的电力。相比之下，基于云的数字神经网络需要千瓦。受大自然的启发，我们开发了NASP：神经形态模拟信号处理器，它模拟了大脑的功能和效率。据估计，欧洲约22%的人口有不同程度的听力损失。总体而言，这项分析表明，全球未解决听力损失的年成本在7500亿至7900亿美元之间，这表明能够解决经济损失并改善福祉的技术存在巨大的市场缺口。我们希望使用我们的NASP技术来改善使用听力设备的人的生活，并认识到它将对那些通过经济高效和简化的解决方案支持他们的困难而降低听力的人产生的影响。目前最先进的数字信号处理(DSP)技术是非智能的、耗电的，并且不能为最终用户提供通信舒适性。我们的研究表明，助听器用户感到不舒服，不愿使用性能低下、需要频繁充电或更换电池的笨重设备。为了应对这些挑战，本项目将开发NASP语音提取芯片。在神经网络的支持下，它将被集成到(任何)听力设备中，并使说话者和听者能够通过APP、电话或面对面进行清晰的交流。我们将创造一个不需要复杂集成的直接模拟接口和足够小的外形尺寸来支持入耳式解决方案。我们的目标是创造新一代信号处理解决方案，并提供一个可行的工具来改善用户体验。

At Polyn Technology, we take inspiration from the human brain. Our eyes pre-process gigabits of visual information and pass 100,000 times less data to the brain, reducing non-essential effort. Thus, the brain, with 80billion neurons, consumes just 20 watts of power. In comparison, cloud-based digital neural networks require kilowatts. Inspired by nature, we have developed NASP: Neuromorphic Analogue Signal Processor, which mimics the brain's function and efficiency.Around 22% of the population of Europe were estimated to have some degree of hearing loss \[www.hear-it.org\]. Overall, this analysis suggests that the annual cost of unaddressed hearing loss is in the range of $750--790bn globally, demonstrating a large market gap for technologies that can address economic loss and improve well-being. We want to use our NASP technology to improve the lives of those using hearing devices and recognise the impact it will have upon those that have reduced hearing by supporting their difficulties with a cost-effective and simplistic solution.Current state-of-the-art Digital Signal Processing (DSP) technologies are non-intelligent, power hungry, and do not provide end-user communication comfort. Our research has indicated that hearing aid users are uncomfortable and unwilling to use bulky devices with low performance that require frequent recharge or battery change.In order to address these challenges, this project will create the NASP Voice Extraction chip. Powered by neural network, it will be integrated into (any) hearing device and enable clear communication for both the speaker and listener, whether by app, phone, or in person. We will create a direct analogue interface with no need for complicated integration and a small enough form-factor to support in-ear solutions.Our aim is to create a new generation of signal processing solutions and to provide a viable tool to improve user experience.