JPH06214597A - Sound processor - Google Patents

Abstract

PURPOSE: To improve the speech processor, regarding special application stimulating an inner ear cochlea-embedded electrode array. CONSTITUTION: This invented speech processor is equipped with a means 8 which receives an electric signal representing a speech signal, a filter means 23 which provides amplitude signals corresponding to the amplitudes of speech signals of frequency channels having plural intervals, a means 27 which selects one or more amplitude signals according to an amplitude degree, a storage means 27 which stores individual stimulating current reaction characteristics, a means 27 which allocates the selected amplitude signal to the stored current reaction characteristics and generates a corresponding current signal, and means 6, 32, and 33 which communicate corresponding current signals to the electrode array, so that the electrodes at positions corresponding to the frequency channels are stimulated with corresponding signals.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice processing device,
In particular, it relates to a speech processing device by a method of processing received acoustic data as well as stimulating an embedded electrode array.

[0002]

General techniques for stimulating embedded arrays are:
US Patent Application No. 4 by Crosby et al.
53293, US Patent Application No. 4207441 by Ricard and others and US Patent Application No. 461.
1598. Such techniques involve implanting an electrode array in the cochlear portion of the inner ear to create hearing, connecting the stimulator to the array by direct or indirect means, and modulating the stimulation according to a signal. Usually included. This signal is usually generated by processing the electrical output of the microphone in a predetermined format.

Many known processing techniques focus on utilizing models of how hearing is detected by the brain in response to specific stimuli. Thus, this data is processed to some extent for the enhancement of specific types of audio information in the stimulation of the electrode array.

[0004]

According to the present invention, however, it has been found that embedding can be usefully stimulated on the basis of improved forms of processing which do not rely on imposing on a particular model in the processed data.

[0005]

According to the invention, the electrical signal corresponding to the received audio signal is processed by the filtering means providing a signal corresponding to the amplitude in the plurality of channels and is selected with the maximum amplitude. A number of such amplitude signals are transmitted to the embedded electrode array and used to modulate. Preferably, the array is stimulated at a constant rate and the stimuli are delivered simultaneously but not simultaneously.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A detailed description of the present invention will be given below with reference to the drawings. Please refer to FIG. This figure shows an overview of the device for stimulation of an array of electrode pairs according to the processed signal.

The electrode array 1 is a reception stimulation unit (RS
U) 3 is connected via cable 2 and is embedded in the cochlea. The entire implantable device may be conventional, such as the cochlear mini system 22.

The embedded device receives the control signal and receives power from an external audio processing unit, preferably an RF as shown.
(Radio frequency) Received via tuning coil devices 5 and 6. However, any other connection technique such as a transcutaneous connection may be used. The RF tuning coil 6 is connected to the RSU (reception stimulating unit) 3 in a desired sequence, timing,
And a signal modulated by the processor 7 to stimulate the electrodes in the electrode array in amplitude.

The processor 7 instead receives an electrical analog signal from a microphone 8 worn by the user. The present invention relates to the operation of the processor, and in particular to a method of processing an input electrical signal. While the present invention has been described with respect to a cochlear implant device, it is generally used for voice processing, hearing aids, voice recognition, voice synthesis and voice tactile display (especially voice information display by hand pressure pattern for acupuncture patients). It can be emphasized that it can also be applied to devices).

Next, please refer to FIG. This figure shows where the sound received by the microphone 8 produces a corresponding electrical signal. Sensitivity controller 21 provides an adjustable attenuator that allows some degree of ambient sound level. The signal is then pre-amplified by 22 and optionally compressed. The signal is a filter bank 2 consisting of parallel filters tuned to the next adjacent frequency channel.
3 is processed. In the preferred embodiment, 25
There are 16 channels with a center frequency of 0 to 5400 Hz, and the filter bank is a single chip device.
Preferably the filter spacing is linear up to 1650 Hz and above that is a logarithmic curve.

Each channel in the illustrated analog means includes a bandpass filter 24n, then a rectifier 25n, and a lowpass filter 26n to provide an amplitude decision for each channel. Preferably each low pass filter has a cutoff frequency of about 200 Hz. The output frequency from each channel is then digitized. The digitized output is modified by the microprocessor 27 to respond to normal changes in auditory sensitivity with frequency. The set of outputs is multiplied by a set of corresponding coefficients so that at about 400Hz the device sensitivity increases slightly, at higher frequencies it decreases, and at about 4kHz the broad peaks gradually increase in sequence. It

The microprocessor then selects six maximum channel amplitudes at intervals of about 4 msec. Note that this usually does not represent 6 different spectral peaks, so that adjacent channels can separate energy from a single spectral peak. The selected amplitude is then converted into a stimulation current level. As with known devices, the current levels corresponding to the audible threshold and maximum fit level for each configuration of electrode stimulation in a particular patient are empirically determined and stored in memory.
The amplitude is then placed in the individual stimulation area for each implanted electrode set. Another way to convert amplitude to stimulus level is
Alternatively or similarly to the current level, the pulse width is changed. The processor selects the appropriate active electrode for each stimulation pulse according to the channel frequency. Data is 3
RF (radio frequency) tuning coil 6 encoded with 2
Sent by 33.

The microprocessor 27 is also a strong sound control unit 3.
Connected to 1. This strong sound control unit is convenient for the user to use in correlation with the sensitivity control unit 21. Strong sound control unit 3
1 essentially allows the current amplitude level (or pulse width) to be adjusted within a predefined range without affecting the device sensitivity to the input signal.

Usually, a maximum of 16 vertexes of stimulating electrode positions are arranged in order of the pitch of the sound of channel 16 of the filter bank. Preferably, the channel selection technique confirms that the maximum rate of stimulation at any electrode is 250 Hz. It should be emphasized that 6 channels from a 16 channel device is just one configuration, and devices with around 6 channels to filter and around 6 selected channels are included within the invention. Other stimulation ratios and temporal sequences of other stimulation pulses further provide satisfactory or improved performance.

In addition, the present invention provides a digital signal processing (DS
P) means can be used to equip. Preferably this uses a Motorola integrated circuit DSP56001. One digital device is a 128-point radix-2 fast Fourier transform (FF).
T), 0 to 5.85kHz (sampling rate 1
It provides discrete spectral values for 65 different frequency channels linearly spaced (1.7 kHz). This fast Fourier transform (FFT) calculates a long time sound train of a voice waveform sample every 4 msec to 10.9 msec. The calculation of each continuous fast Fourier transform (FFT) therefore comprises, every 6.9 msec, a pre- and sequential sequence of tones.

Prior to the calculation of the Fast Fourier Transform (FFT), this temporal sequence is windowed by forming a function to provide a temporal implementation of the desired spectrum and filter bank. This window function consists of Daniel Windu modified by Kaiser Windu (where θ = π) (tapered sideways in the frequency domain and the top is flat). It is 180 at -3dB point
Provides a filter bandwidth of Hz. Note that the Fast Fourier Transform (FFT) spectrum sample is 91Hz
And thus each second sample is omitted, leaving 32 spectral samples at 182 Hz intervals. The direct current (0 Hz) value is also omitted.

The 32 discrete samples are approximated by the sum of the forces in the next adjacent spectral sample, approximately 16
Is reduced to the spectrum of. The resulting 16 channel filter bank is arranged such that a minimum of 8 channels have equal bandwidth and are linearly spaced. Up to 8 channels are arranged to have a near logarithmic increase in filter bandwidth and spacing. Preferably, the center frequency of 16 filter channels is 27
4, 457, 640, 823, 1005, 1188, 1
371, 1554, 1828, 2194, 2559, 2
925, 3382, 3747, 4296 and 5118Hz
Is.

Each of the 16 spectral channels is assigned to a single stimulating electrode location in a tonal arrangement. 6
One electrode is stimulated during each analysis period (ie every 4 msec).
The six stimulated electrodes are selected based on the instantaneous amplitudes of the 16 spectral components. The six largest spectral components in each analysis period are selected in analog form. The amplitudes of the six selected spectral components are transformed using an intensity-producing power function and then placed in the active area of the stimulation electrode. The 6 stimuli are ordered from maximum to minimum amplitude and continue rapidly every 4 msec and are delivered to the implant.

[0019]

It should be noted that this technique yields results at a relatively constant ratio to the stimulus compared to many prior art techniques, however experimental evidence is that the user can It is to be pointed out that speech recognition by using the processor is at least as good or better than speech recognition by using other processors.

It should also be noted that the device provides improved performance over the prior art over silence, since it does not assume a voice-out received voice.

The selection of related components and circuit design will be described below. In order to obtain a satisfactory result in the present invention, it is necessary to pay attention to the following points. Microphone characteristics and frequency response affect the quality of the processor's signal input, and certain modifications of the equalization technique can improve performance.

When the microphone and transducing coil are mounted on a regular headphone, attention should be paid to RF (radio frequency) interference between them. This combination creates a component in the acoustic domain. The signal to noise ratio can be improved by including a suitable preamplifier (eg 40 dB gain) in the headphones or microphone at the end of the cable. You need to make sure that the power supply is properly tuned to avoid ripple and noise. Other devices and modifications are possible within the spirit and scope of the invention.

[Brief description of drawings]

FIG. 1 is a block diagram showing an implantable nervous system stimulating device.

FIG. 2 is a voice processing block diagram showing a voice processing device according to the present invention.

[Explanation of symbols]

 1 ... Electrode array 3 ... Receiving stimulation unit (RSU) 5, 6 ... RF tuning coil device 7 ... Processor 8 ... Microphone 21 ... Sensitivity control section 22 ... Preamplifier 23 ... Filter bank 27 ... Microprocessor 31 ... Strong sound control section

Claims (2)

[Claims] 1. Means (8) for receiving an electrical signal representative of an audio signal, filter means (23) for providing an amplitude signal corresponding to the amplitude of the audio signal in a plurality of spaced frequency channels, An audio processing device comprising means (27) for selecting one or more of the amplitude signals according to the width and means (27) for generating one or more output signals corresponding to the selection signals. 2. Means (8) for receiving an electrical signal representative of an audio signal; filter means (23) for providing an amplitude signal corresponding to the amplitude of the audio signal in a plurality of spaced frequency channels; Means (27) for selecting one or more of the amplitude signals according to the amplitude, storage means (27) for storing individual stimulation current response characteristics, and the selected amplitude signals for the stored current response characteristics. And communicating the corresponding current signal to the electrode array so that electrodes at positions corresponding to the frequency channel are stimulated with the corresponding current signal. A sound processing device for generating a stimulus signal to an electrode array comprising means (6, 32, 33).