| The skinhearing system is a sensory substitution system that uses tactile substitute the auditory sense whose core is transmitting sound information through the sensory nerve to the brain and completing the neuroplasticity through speech recognition training.Skinhearing technology firstly extracts the spectral information of speech,and then divides the information into multiple channels and stimulates the sensory nerves of the skin through multiple microcircuits so as to realize the conversion of speech information into sensory information.The study of skinhearing speech recognition method has found that there are still existing the problems that the adjustment of the stimulation channel number is not flexible enough,the sensory thresholds is affected by the variation of the stimulation current frequency,and the output amplitude of the low-frequency channels is lower than the expected value.In this paper,in order to solve the problems of skinhearing,a multi-channel experimental platform of skinhearing is designed and developed,and the speech processing strategy of skinhearing is improved based on CIS algorithm.The specific work is summarized as follows:1)Design the skinhearing multi-channel experimental platform,including the aspect of hardware,such as USB control module,DAC module,and the aspect of software,such as the USB device’s firmware,the USB device’s driver and applications.In hardware design,a 48-channel experiment platform is designed by using the chip of Cypress EZ-USB FX3 USB 3.0 and digital-analog conversion chip of CS4384.The data transmission between USB and DAC is implemented by GPIF Ⅱ technology.While in software design,the firmware program initializes the USB device and controls to receive and transmit the USB data packets to realize the simultaneous working of multiple DACs;The application program uses the Lab VIEW program to realize the speech acquisition and processing,and transmits the speech to the USB bus.2)In the specific design process of the experimental platform,the required interface frequencies for the synchronous transmission of GPIF II and DAC data is calculated according to the digital interface format of CS4384,and the data format of 48 channels sampled once by the DAC is defined,and two DMA channels are implied at the same time used to cache data to improve system refresh rate.3)Through the study of skinhearing technology and Continuous Interleaved Sampling(CIS),aiming at the problems of the influence of the frequency of the output current on the sensory thresholds and the low output amplitude of the low frequency channels,this paper proposes an improved method of skinhearing speech processing based on the CIS algorithm.The improved strategy uses the short-time Fourier transform(STFT)to realize the CIS algorithm.The signal of each channel is amplitude-modulated and then output.The LabVIEW program designs and realizes an improved skinhearing speech processing strategy,including signal pre-emphasis,STFT,accumulation of short-term energy,compression coding and amplitude modulation.The paper completed the software and hardware design of the skinhearing multi-channel experimental platform,proposed and implemented the improved skinhearing speech processing strategy based on the CIS algorithm.The results show that the experimental platform can achieve 48-channel signal output with flexible adjustment of the stimulation channel,high refresh rate and low average delay.The improved skinhearing speech processing strategy based on CIS algorithm unifies the frequency of the output current of all the channels and increases the amplitude of the output of the low-frequency channels.Moreover,the strategy helps speech recognition and the super-multiple output channels of the experimental platform can recognize the lingua-palatal phonemes more effectively. |
