Speech-in-Noise & Directional Mics Directional microphone technology wasinvented by the military during WorldWar II. It’s concept was refined andminiaturized into hearing instrumentsin the late 1960’s.
Speech-in-Noise & Directional MicsHearing instruments designed with thistechnology would, theoretically, provide abetter listening experience for the hearingimpaired. However, the early microphonesproduced so much of their own noise whenoperating that they were of limited utility.
Speech-in-Noise & Directional Mics As HI dispensing professionals, there aretwo things we must do for the hearingimpaired patient/client:1.Improve audibility2.Improve signal-to-noise ratio
Speech-in-Noise & Directional Mics The amplified gain of hearing instruments isused to provide the audibility of sound forthe hearing impaired. However, amplified audibility does notalways provide a greater opportunity forspeech recognition in difficult listeningenvironments i.e. speech-in-noise ability.
Speech-in-Noise & Directional Mics Directional microphones have proven toprovide a greater opportunity of speechrecognition in noisy environments. In fact, refinements in the 1990’s providedfor the remote control of the directionalmicrophone ability by the patient/client.
Speech-in-Noise & Directional MicsDIRECTIONAL MICROPHONES They do not increase the sound input fromthe front of the listener—they reduce theinput of the sound behind them. It is hoped that the desired input signal is infront of the patient/client and not behindthem!
Speech-in-Noise & Directional MicsConventional•1 mic, 2 ports•on BTEs only•often no “on/off”Newer•2 omni mics•found on ITEs•routinely “on/off”
In Any MicrophoneSound Moves DiaphragmDiaphramSource
DiaphramSource SourceSounds Hitting Both SidesCancel Each Other Out
Directional Microphone FunctionWhen Sounds Come From Front…)))The Diaphragm moves))FilterRearFront Direction of incoming sound
Speech-in-Noise & Directional MicsDIRECTIONAL MICROPHONESWith this fundamental front—back operation,it was found that the greater the separationbetween the two microphone openings(ports), the more effective the speech-in-noisebecame.
Speech-in-Noise & Directional MicsDIRECTIONAL MICROPHONESThey do not restorenormal hearingability—theystimulate theresidual ability ofthe patient/client.HEARING LOSS IS PERMANENT!
Speech-in-Noise & Directional MicsDIRECTIONAL MICROPHONESThere is currently new and moreexpensive directional technology whichinvolves two separate microphoneswithin the hearing instrument i.e. “dualmicrophone processing”.
Speech-in-Noise & Directional MicsDIRECTIONAL MICROPHONES ANSI has established a standard formeasurement of their effectiveness. It iscalculated as the Directivity Index (DI). The greater the DI, the more effective it isregarding the separation of the signal-to-noise.
Speech-in-Noise & Directional MicsPOLAR PLOTS The directional performance graphs/chartsare represented as Polar Plots. Directional performance polar plots begin byrepresenting a Directivity Index (DI) aszero—(no DI). It’s graphical appearance is represented as aperfect 360 degree circle on the polar plot.
Speech-in-Noise & Directional MicsPOLAR PLOTS An increased DI number represents theeffectiveness of directional microphoneactivity. The DI number increases, as polar plotsreveal null points (areas where sound isreduced in intensity).
Speech-in-Noise & Directional MicsPOLAR PLOTS These reduced intensity areas arerepresented as indentations into the perfectcircle. These indentations are referred to as thenull areas of intensity.
Speech-in-Noise & Directional MicsDIRECTIONAL MICROPHONES Most of today’s typical directionalmicrophones generally represent aDirectivity Index (DI) of about 5dB to6dB.
Speech-in-Noise & Directional MicsDIRECTIONAL MICROPHONES Their effectiveness is, of course, influencedby the frequencies it/they receive. In other words the directivity index (DI)varies by the different input frequencies. In fact, many algorithms for today’sinstruments have automatic reduction of lowfrequency inputs when directionalmicrophone activity is initiated.
Speech-in-Noise & Directional MicsDI and Frequencies The most important frequencies forunderstanding speech are: 1kHz - 4kHz (themost important of these is 2kHz). One can simply take average DI’s of 4polar plots to calculate the overall microphoneDI for hearing instrument performance.
Speech-in-Noise & Directional MicsARTICULATION INDEX To determine the best performance forspeech understanding, dots withinthe “speech banana” were created. This was to identify the most effectiveinput frequencies to be “processed” bydirectional microphones.
Speech-in-Noise & Directional MicsARTICULATION INDEX There are one hundred dots, with each dotrepresenting a one percent contributiontowards speech intelligibility. You will notice that most of the dots arelocated within the higher frequencies of the“speech banana”.
Speech-in-Noise & Directional MicsARTICULATION INDEX
Speech-in-Noise & Directional MicsSPEECH UNDERSTANDING IN NOISE Normal hearing allows for an understandingof speech-in-noise ability to be achieved fiftypercent of the time; this should occur whenthe speech signal intensity is equal to thenoise signal intensity.
Speech-in-Noise & Directional MicsSPEECH UNDERSTANDING IN NOISEIn 1997, Meade Killion’s research discoveredthat with every one decibel of improvement ofthe speech signal over the noise signal, a tenpercent improvement was realized for theability to better understand speech in noise.
Speech-in-Noise & Directional MicsCertainly, the five to six decibelimprovement in the signal-to-noise ratioexhibited by today’s directionalmicrophones, can reflect a fifty to sixtypercent expected improvement inperformance for the patient/client—ifthey have residual hearing ability tostimulate.
• Directional microphones objectivelyimprove speech/noise performance• Digital noise reduction subjectivelyenhances comfort in noiseSolutions for speech understanding inNOISESpeech-in-Noise & Directional Mics
Adaptive DirectionalityThis technology is designed for the use of twin omni-directional microphones. It can:• Automatically shift from Omni to Dmic (Dependingupon listening environment).• Automatically switch among various polar plots(Depending on listening situation).• Shift polar plot nulls to origin of noise (Dependingon noise source direction).NOTE: It is not necessarily statistically better…Buthas advantages for those with poor manual dexterityand those who cannot tell when to use a feature or whatfeature should be used.
Speech-in-Noise & Directional Mics Adaptive directionality created withdual microphones has not shown anyfurther improvement in the SNR Multiple microphone (more than two)technology can provide greater than thefive to six decibels of signal-to-noiseratio improvement. However, is it practical?
Beam forming:Making Directional Mics Better YetMics with more than 2 ports•eg. 3 or more micsThis is Killion’s ArrayMicTM•heart is in right place•DI’s are about 7-10dB!Photo providedCourtesy M. Killion
Spectral Subtraction1. The spectrum of speech & noise together is received2. During pauses in conversation, the spectrum of noise isestimated.3. The spectrum of speech & noise is subtracted by thespectrum of noise only.4. Theoretically, this leaves just the spectrum of speech.Problem:• The wide noise spectrum intersects with the speechspectrum.• Thus, removing noise removes some of speech.
5075Frequency (kHz).1 1. 101. Speech Plus Noise Spectrum…Speech with its 6db/octave roll-offNarrow bands of noisedBSPL
Phase Cancellation1. Exact time waveform of noise is measured.2. Inverted noise phase added to the original noisewaveform cancels the noise.This phase+Oppositephase=
Because in the headphones:• Speech is sent directly to the eardrum fromthe headphone.• Noise is sampled by the microphone outsideof the headphone.Digital hearing aids:• Do not have this luxury.• As both the speech & noise inputs arepicked up by outside microphone.Phase Cancellation used innoise reduction headphones…Why not in hearing aids?
Why Phase Cancellation Can Work in HeadphonesBut Not in Hearing AidsSpeech entersdirectlyfrom headphoneNoise from outsideleaks into earcanal and mixeswith speechNoise from outsidepicked up bymicrophone andinverted in phase THIS CANCELS OUT THE NOISE
SpectralEnhancement1. A digital algorithm detects spectral speech cues innoise such as vowel formants or high-frequencysibilants.2. The algorithm deliberately enhances or amplifies thesespectral speech cues. This is just a different approachfrom noise reduction.
The challenge for SpectralEnhancement, is the high-frequencyconsonants:In noise:• The valleys b/w peaks of speech are filledw/noise the peaks are thus less prominent• The low-frequency vowels are more intense,so these still stand out--it is easier to enhancethese.
The challenge for Spectral Enhancement,is the high-frequency consonantsIn noise:The valleys b/w peaks of speech are filled w/noisepeaks thus, they become less prominent.The low-frequency vowels are more intenseSo, these still stand out. It is easier to enhance these.
Speech Synthesis1. The digital algorithm detects spectral speech cues innoise.2. Once a particular speech sound is detected itthen adds a similar synthesized speech sound.NOTE: It requires a “stored collection” of speech sounds.This may result into:• A difficulty to digitally recognize speech sounds.• Overwhelming complexity.• A synthesized speech sound which can soundunnatural.
Today’s Digital Hearing Aids Use:A weak form of Spectral Subtraction is an amplitudemodulation approach; this sometimes is a combo offrequency and “time-of-duration” modulation.By subtracting the noise spectrum from the noise +speech spectrum, it may remove too much speech.DSP algorithms have characterized waveforms. In eachchannel the noise has a fairly flat waveform. Over timethe speech waveform fluctuates rapidly.If noise is sensed within a channel,then gain is reduced some 5-20dB.
For Speech:Mean Intensity is Not in Middle of Range255075Frequency (kHz).1 1K. 10Long Term Average Speech Spectrum (LTASS)dBSPLTHIS IS BECAUSE SPEECH HAS ABNORMAL DISTRIBUTION OF INTENSITY
Noise ReductionMost Digital Hearing Aids Use It...Sounds that don’t change in intensity are reducedSounds that do change in intensity (speech) are not reduced
How often(eg. %)the intensityof thesoundis atsomeparticulardB levelSpeech (single talker)Decibel LevelNoiseSpeech has an odd distribution of intensity
125 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000Gain(dB)Noise Reduction with OneChannel…J m d b iuazvrn onge lph gchshkfsth
125 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000Gain(dB)Noise Reduction with OneChannel…Reduces Gain Over All Speech Hz’s!All Speech Sounds DropJ m d b iuazvrn onge lph gchshkfsth
125 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000Gain(dB) Noise Reduction withTwo Channels.Isn’t much better?J m d b iuazvrn onge lph gchshkfsth
125 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000Gain(dB)Noise reduction wouldreduce half of the gain overthe Speech frequencies!Vowels Would DropConsonants Would NotJ m d b iuazvrn onge lph gchshkfsth
125 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 70001 2 3 4 5 6 7 8 9 10 11 12 13 14 15Center Hz of Each BandGain(dB)Noise reduction with lots ofbands/channels...J m d b iuazvrn onge lph gchshkfsth
125 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 70001 2 3 4 5 6 7 8 9 10 11 12 13 14 15Center Hz of Each BandGain(dB)Would reduce gain over smaller Hz regionsJ m duazvrnnge lphgchsh kfsthob i
Digital Noise Reduction+Directional MicrophonesNoise reduction algorithms•give subjective comfort to clientDirectional microphones•gives objective improvement in speech receptionTogether they make a good team!•a twin-headed approach to speech-in-noise