Table Of Contents:

1   Introduction

2   Equipment
     2.1   Workstation
     2.2   Headphones and Speakers
     2.3   Optech Laser Range Finder (LRF)
          2.3.1   Operation of the LRF
               2.3.1.1   Power
               2.3.1.2   Turning the Unit on (Obtaining a Reading)
               2.3.1.3   LRF Messages
               2.3.1.4   Remote Terminal Connection
               2.3.1.5   Control Commands Via a Serial Connection
               2.3.1.6   Description of the Serial Port Data Transmitted
               2.3.1.7   LRF Specifications

3   Source Code
     3.1   lrf_functions.h
     3.2   lrf_functions.c
          3.2.1   intSerialPortInit(char *serialPortName)
          3.2.2.   int sendCommand(char *cmnd, int cmdLength)
          3.2.3.   int fdHasCharacters(int fileDes)
          3.2.4.   int getReading(unsigned char *buffer)
          3.2.5.   float processReading(unsigned char* reading)
          3.2.6    void lrfSetUp()
          3.2.7    double calcAvgDistance(double *readings, int numOfReadings)
          3.2.8    void error(char * errorMsg)
     3.3   lrf_audiov2.h
     3.4.   lrf_audiov2.c
          3.4.1.   int audioInit(char *audioDevice, char *audioCntrlDevice)
          3.4.2    int sendAudio(short *buffer, int numOfSamples)
          3.4.3    void tableInit(short *table[], int varyMagnitude)
          3.4.4    short *makeWave(double frequency, double magnitude)
          3.4.5    short *makeDepthSignal(in numOfSamples)
          3.4.6    short *createNoise(int numOfSamples, double excitation)
     3.5    lrf_main.c (Executable name:  lrf)
     3.6    lrfCont.c (Executable name:  lrfCont)
     3.7    lrf_piano.c (Executable name:  lrfPiano)
     3.8     Makefiles
     3.9    Additional Code Written
          3.9.1    playTone.c (Executable name:  tone frequency [magnitude])
          3.9.2    cosLookUp.c
          3.9.3    Time Taken to Execute Portions of Code
     3.10   Program Requirements

     4   Sound Output
          4.1    Sample Rate
          4.2    Audio Encoding and Sample Precision
          4.3    Distance to Frequency Mappings
               4.3.1    Linear Mapping
               4.3.2    Inverse Square Mapping
               4.3.3    Logarithmic Mapping
               4.3.4    Piano Keyboard Frequencies
     4.4    Distance to Mangnitude Mapping
          4.4.1    Inverse Square Law Mapping
     4.5    Generation of the Samples
          4.5.1    Output of a Continuous Tone
     4.6    Real-Time Sample Generation vs. Wave look-up Table
     4.7    Depth Discontinuity
          4.7.1    Summation of Sine Waves
          4.7.2    Triangular and Pulse Signals
               4.7.2.1    Probability:  The Triangular Distribution
               4.7.2.2    Triangular Formula
          4.7.3    Using and Average of the 'N' Previous Distances

     5    Internet Sites Visited
          5.1     Related Research (past and present)
               5.1.1    Obstacle Detectors
               5.1.2    Orientation and navigational Aids
               5.1.3    Environmental Aids
               5.1.4    Virtual Sound
          5.2    Commercial Products
          5.3    MIDI Sites
          5.4    Miscellaneous Sites
               5.4.1    Linksa to Related Sites
          5.5    Newsgroups

     6    Suggestions and Conclusions
          6.1    Ideas for Future Continuation of the Project
               6.1.1    Use of MIDI to Generate Sound
               6.1.2    Use of a Macintosh over the SUN Workstation
               6.1.3    Eliminating the Effects of the "Curves of Equal loudness"
               6.1.4    New 'Power' and 'Data Out' Connectors
          6.2    Conclusions
          6.3    Acknowledgements

     7    References
          7.1    Bibliography
          7.2    Source Code Refernces

1. INTRODUCTION:

Distance measurements could  be easily obtained using the Optech Laser Range Finder (a high accuracy infrared ranging device).   Therefore goal of my project was to obtain these distance readings and convert them into audible sound ultimately leading to a product which will aid the visually impaired with their travel and mobility.

The following pages describe the work I have completed during this four month summer period.

 2. EQUIPMENT:

2.1 WORKSTATION

Either the SUN Ultra 1 or the SUN SparcStation 5 workstation was used. The software written for this project is compatible with both.  Each workstation contains two serial ports (A and B) and each serial port contains a 25 pin female RS-232-C compatible connector.  The serial ports can be accessed with the following names:

• Port A:  "dev/ttya"
• Port B:  "dev/ttyb"

2.2 HEADPHONES AND SPEAKER

The sound may be output with the workstation's built-in-speaker or with headphones.  Panasonic XBS headphones were used and allowed for a greater quality of sound compared to the built-in-speaker especially with the frequencies above 4kHz.
 

2.3 OPTECH LASER RANGE FINDER (LRF)

Optech Laser Range Finger (LRF) model G-150 is used to obtain the distance readings.  "The Rangefinder contains an infrared laser source emitting infrared radiation invisible to the human eye (890 nanometeres) and should not be viewed directly with the eye or with optical devices when the laser is firing ." (Model G-150 owner's manual).

2.3.1  OPERATION

2.3.1.1 POWER

2.3.1.2 TURNING THE UNIT ON (OBTAINING A READING)

1. Manually:
By pressing the  START button to obtain a single measurement (Single   update mode) or can be set to take continuous readings (Continuous update   mode) by installing the DATA IN adapter plug.  Plugging in the adapter plug   prior to turning the unit ON results in readings being taken immediately    following power ON without pressing the START button.
 
2.  Computer:
A computer generated "start" pulse signal (any positive going pulse) can be sent to the LRF by a remote computer to initiate a reading (Single update mode).  Continuous update mode can also be achieved by sending the appropriate command (see table 2.2 for a list of the available commands).
 

2.3.1.3 LRF MESSAGES

    READY        Indicates the unit has passed its self-test and can begin taking reading.

    *                    Replaces the decimal point in a range reading to indicate the battery  is low
                          (below9.5V).  Appears on the display as well as the remote
                          terminal.

    <                   Appears to the left of the range reading indicating the over 12.5% of the shots were
                         invalid (drop-outs).  This is ignored in my present  program implementation.

    <<<<             Appears in place of a reading (both on the display and the remote terminal)
                         indicating an accurate reading could not be taken.

    ERROR        Appears on the display when there is a problem with the unit..

2.  BEEPER

• 1 SHORT BEEP        Normal reading obtained.
• 2 SHORT BEEPS      Valid reading however, more than 12.5% shots were drop outs. .                                  A '<' also appears on the display.
• 1 LONG BEEP           Unable to take an accurate reading.  '<<<<' also appears.

2.3.1.4 REMOTE TERMINAL CONNECTION

Several of the solder connections on the D9S became loose and were re-soldered.  To avoid any wiring mix-up, table 2.1 below shows the correspondence between colour wire coming from the six pin circular connection and the pin number of the D9S connector

2.3.1.5 CONTROL COMMANDS VIA A SERIAL CONNECTION

The commands can be sent at any time and in any order however, all the commands must contain a Carriage Return (CR) as their last character.

In order to process each command, the LRF requires a certain time delay.  According to the Optech manual, this delay should be 200usec, however, I discovered the delay time required is about 1ms.
This does not affect the proper functioning of my program since commands are issued once only prior to obtaining any readings.

2.3.1.6 DESCRIPTION OF THE SERIAL PORT DATA TRANSMITTED

VALID READING:
Three characters representing the whole portion of the reading, a character representing an ASCII period
followed by two characters representing the decimal portion.  When the whole portion of the reading requires less than three digits, the character(s) not used are assigned SP.

DROP-OUT READING:
four ASCII characters representing '(', followed by two SP characters.

NOTE:    With a terminal connection to the LRF, upon turning the LRF ON, the ASCII characters
                representing the string 'READY' are sent to the terminal.

2.3.1.7 LRF SPECIFICATIONS

• Distance readings can range from 0.2m to 100m.  For my purposes, a range between 0.5m and 15m was used.  I have also found trying to obtain a distance reading less than 0.2m results in a drop-out (indicated by '<<<<').

• An accuracy of ?10cm is achieved between the temperature ranges of -20oC and 50oC.

• The FAST/SLOW switch sets the time taken to obtain a reading by controlling the number of shots being averaged (each distance reading is an average of all the shots fired in order to reduce random errors).
• Two readings/second are obtained in the SLOW setting (1000 shots/reading), while eight readings/second in the FAST setting (150 shots/reading).  With a serial connection many more shots/reading settings are permitted.
• With a serial connection, during power ON the unit will respond to the SLOW/FAST setting, however, this can be changed by sending the desired shots/reading at any time.  Valid settings can range from  1-9999 and must be followed by a Carriage Return ("\r").  For example, for a setting of 150 shots/reading, "150\r" should be sent.
•  Decreasing the number of shots/reading decreases the time to obtain a reading, however, it also reduces the accuracy of the distance readings.

• The LRF was used primarily during the day although I have experimented with it during the night with a minimum amount of available light and have found it works fine.
• I have not had the chance to use the LRF during a snowfall or rainfall although according to the owner?s manual, it should work fine.
• Windows pose a problem with object recognition since the LRF will not measure the distance to a window. Windows do not reflect the signal instead the signal passes through the window and a measurement of an object beyond the window is taken.

• "Model G150 Rangefinder User Manual" Version 1.1, May 1991
• "Model G150 Release Notes" June 11 1992

Bob Karwal
Optech Systems
701 Petrolia Road
North York, Ontario
M3J 2N6

Tel: (416)661 - 5904
Fax: (416)661 - 4168

3. SOURCE CODE

3.1 lrf_functions.h

3.2 lrf_functions.c

3.2.1 intSerialPortInit(char *serialPortName)

• DATA SPEED:                                              9600 Baud
• NUMBER OF DATA BYTES:                        8
• NUMBER OF STOP BITS:                             1
• PARITY:                                                       NONE
• MIN NUMBER OF BYTES TO APPEAR:       0
• TIME TO WAIT FOR DATA:                        0

Further documentation regarding the serial port can be found in the manual pages under the following titles:

• termios
• ioctl
• fcntl
• open(2)

3.2.2 int sendCommand(char *cmnd, int cmdLength)

3.2.3 int fdHasCharacters(int fileDes)

3.2.4 int getReading(unsigned char *buffer)

3.2.5 float processReading(unsigned char* reading)

• Determine whether the reading is 'out-of-range'.  Out-of-range readings cannot be processed and result in a return of 0.0.
• Process the digits to the left of the decimal point (whole number portion).  Any reading without a decimal point signifies an error and results in a return of 0.0.
• Process the digits to the right of the decimal point (fraction portion).
• When the converted reading value is within the acceptable range, it is returned otherwise either MIN_DISTANCE (when the reading is below MIN_DISTANCE ) or MAX_DISTANCE  (when the reading is greater than the MAX_DISTANCE) is returned.

3.2.6 void lrfSetUp()

3.2.7 double calcAvgDistance(double *readings, int numOfReadings)

3.2.8 void error(char * errorMsg)

3.3 lrf_audiov2.h

3.4 lrf_audiov2.c

3.4.1 Int audioInit(char *audioDevice, char *audioCntrlDevice)

audioInfo.play.port = AUDIO_HEADPHONE

• Controls the output of the audio port.  Output can be sent to either of the following:
• . Workstation built in speaker (AUDIO_SPEAKER).
•  Headphone jack (AUDIO_HEADPHONE), to allow the use of headphones.

• audioInfo.play.balance = AUDIO_MID_BALANCE
• Control the volume between the left and right channels.
• Values between 0 and 64 are accepted.
• When set between AUDIO_LEFT_BALANCE (0) and AUDIO_MID_BALANCE (32),  the right channel will be reduced in proportion to the balance value.
• When set between AUDIO_MID_BALANCE and AUDIO_RIGHT_BALANCE (64),  the left channel is reduced in proportion to the balance value.

• audioInfo.play.gain = AUDIO_GAIN
• Output volume control.
• Values between 0 (AUDIO_MIN_GAIN) and 255 AUDIO_MAX_GAIN are accepted.
• see lrf_functions.h for the definition of AUDIO_GAIN.

• audioInfo.play.sample_rate = FREQ_SAMPLING_RATE;
• Sampling frequency of the audio data (in samples per second).
• Currently set to 22050Hz.
• See lrf_functions.h for the definition of FREQ_SAMPLING_RATE.
• See 'audiocs' in the manual pages for the allowable sample rates.
 
• audioInfo.play.encoding = AUDIO_ENCODING_LINEAR;
• Audio data representation can be set to the following:
• 1. Linear PCM (AUDIO_ENCODING_LINEAR).

2. u-Law (AUDIO_ENCODING_ULAW).
3. A-Law (AUDIO_ENCODING_ALAW).
• Currently use Linear PCM 16-bit encoding (see section 4).

• audioInfo.play.precision = SAMPLE_PRECISION;
• Number of bits used to represent each audio sample.
• Allowable sample values are 8, 16 or 32.
• Currently set to 16 bits allowing for sample values between -32768 and 32767.
• See lrf_functions.h for the definition of SAMPLE_PRECISION.

A 1 will be returned upon successful opening and setting of the audio port and 0 if any errors are encountered.

I have made limited use of the audio device.  Many more settings and uses are possible.  For further information regarding the audio device, see the manual pages under the following:

• audio
• audiocs

3.4.2 int sendAudio(short *buffer, int numOfSamples)

3.4.3 voidTableInit(short *table[], int varyMagnitude)

3.4.4 short *makeWave(double frequency, double magnitude)

3.4.5 short *makeDepthSignal(in numOfSamples)

3.4.6 short * createNoise(int numOfSamples, double excitation)

3.5 lrf_main.c (Executable name:  lrf)

Prior to program execution, the LRF must be turned ON and in the "READY" mode.  Should the LRF be turned OFF, the program must be stopped and re-executed.  "CNTRL-C" to stop program execution.

3.6 lrfCont.c (Executable name:  lrfCont)

3.7 lrf_piano.c (Executable name:  lrfPiano)

3.8 Makefiles

• Version 1:    makefile
• Version2:     makefileContinuous
• Version3:     makefilePiano

• lrf_functions.o
• lrf_audiov2.o
• libaudio.a

<unistd.h>            <ctype.h>            <errno.h>            <sys/termios.h>
<fcntl.h>              <time.h>              <sys/filio.h>          <stropts.h>
<stdio.h>             <sys/audioio.h>    <sys/file.h>          <sys/ioctl.h>
<termios.h>         <math.h>             <stdlib.h>
"audio_device.h"  "audio_hdr.h"

3.9 ADDITIONAL CODE WRITTEN

3.9.1 playTone.c (Executable name:  tone frequency [magnitude])

The makefile for this program is makefileTone.

3.9.2 cosLookUp.c

3.9.3 TIME TAKEN TO EXECUTE PORTIONS OF CODE

The UNIX function 'gettimeofday()' returns the current time expressed in  elapsed  seconds  and microseconds since 00:00 Universal Co-ordinated Time, January 1, 1970 (manual pages - gettimeofday).  To determine the time required to execute a segment of code, the time of day is obtained just before the code segment is entered and subtracted from the time of day obtained just after the execution of the code segment as shown below:

#include <time.h>

struct timeval t1, t2;
int *temp = NULL;

gettimeofday (&t1, temp);

** CODE TO BE TIMED IS INSERTED HERE **

gettimeofday (&t2, temp);

printf ("Execution time:  %5.3f s\n",
   (float) (t2.tv_sec - t1.tv_sec) + ( (float) (t2.tv_usec - t1.tv_usec) / 1000000.0 ) );

As shown above, the function gettimeofday() requires two parameters, the first being a timeval structure (defined in time.h), and the second a pointer (of any type), to NULL.  Execution time will be displayed in seconds.

The above code proved to be very useful to determine the time required to obtain a reading, process and output the samples corresponding to it.  For example, in the FAST setting (150 shots/reading), the LRF will take a measurement every 0.125s.  The time required to process a reading was determined to be 0.091s (using the wave-table), ensuring proper timing.

3.10 PROGRAM REQUIREMENTS

• The LRF control commands (including the command to initiate readings), are sent once only.  As a result, the LRF must be powered ON and in the 'READY' mode prior to program execution otherwise readings will not be initiated.
• In the case power to the LRF is removed during program execution,  the program must be halted and re-executed.
• CNTRL-C must be used to stop program execution.

4. SOUND OUTPUT

The maximum frequency used in all the mappings is 6000Hz.  Frequencies above this level, using the present equipment, results in poor sound quality.  Moreover, human perception of higher frequencies  steadily deteriorates after age 30  (?????) and it has been estimated that 65% of visually impaired and partially sighted people are over 70 years of age (Lacey, 1997).

In addition to varying the frequency with distance, the user may choose to alter the magnitude of the wave inversely with distance, giving an additional cue to object location.

4.1 SAMPLE RATE

Another consideration to limiting the sampling frequency is memory space.  Higher sampling rates require a greater number of samples.  However, insufficient memory space has not been a problem, whether generating samples in real-time (in which 4000 samples at 2 bytes each is used at any time - see section 4.6), or using a sample look-up table (where 145 waves of 4000 samples at 2 bytes each are stored in the table - see section 4.6).

4.2 AUDIO ENCODING AND SAMPLING PRECISION

The cause of the above scaling problem has yet to be determined, however, it was eliminated by switching over to Linear Pulse Code Modulation (PCM) encoding.

PCM is an uncompressed audio format in which sample values are directly proportional to audio signal voltages.  According to the Sun Microcomputer Systems manual, each sample is a 2's complement number that represents a positive or negative amplitude (Sun Microcomputer Systems manual pages - 'audio').  Sample size was increased to 16-bit (either 8 or 16-bit is allowed with PCM), allowing for a closer approximation of the signal. In contrast to  the u-law encoding, multiplying the samples by a value less than 0.5 caused a decrease in the magnitude  (multiplying the samples by 0.01 or less caused the tone to be non-audible).  In addition, multiplying the samples by a constant greater than 1 and up to approximately 20, caused an increase in magnitude.  An increase in magnitude was imperceptible using values above 20 (e.g. 20 gave the same effect as 100).

4.3 DISTANCE TO FREQUENCY MAPPINGS

4.3.1 LINEAR MAPPING

• Divide the distance reading by distance increment (0.1m).
• Multiply the value obtained above by the fixed frequency per distance increment.

4.3.2 INVERSE SQUARE LAW MAPPING

• FREQUENCY = k * ((1/distance²) + c)
• FREQUENCY = 1333.33 * ((1/distance²) + 0.5)

Using the inverse square law was a marked improvement over linear mapping, however, as shown in table 4.1 of the following page, frequencies changed drastically for a very small distance range (distances close to the user), and changed only slightly for the greater distance readings.

Although this mapping was very pleasant to listen to (when distances were small and a change in frequency could be perceived), it was not used, largely due to the minimal change in frequency for the farther distances.  Perhaps in the future, this mapping could be used for very close distances, while another mapping could be used for farther distances.

4.3.3 LOGARITHMIC MAPPING

Subtracting the minimum distance from each distance reading to obtain the exponent enables the maximum frequency to be used as the frequency constant. The exponent is then multiplied by a constant ('a'), to increase the overall frequency for each reading.  For example, without this constant, a distance reading of 10.00m would result in a non-audible frequency of 0.45Hz, whereas using the constant a = 0.15 results in a frequency of 1443.05Hz.  Both constants 'c' and 'a' were arbitrarily chosen.  Various other values of  'c' and 'a' were tested (see figure 4.1 and 4.2 for curves).  Using smaller values for 'c' results in a smaller frequency range (see table 4.3).
 

4.3.4 PIANO KEYBOARD FREQUENCIES

As shown below, The frequencies of the piano keyboard range from 27.50Hz to 4186Hz. This follows the scale of equal temperament in which every octave (a 2:1 change in frequency), is divided into 12 equal intervals allowing for the frequency of adjacent notes to differ by a factor of 1.05946 (twelfth root of two.

In order to use this mapping, the frequencies corresponding to the piano keyboard are stored in a table and mapped linearly to distance.  Upon obtaining a distance reading, the corresponding frequency was looked-up, the samples were generated and output, (samples were created in real-time). To determine the appropriate frequency for a distance reading, the following algorithm is used:

• Divide the distance range into 88 sections representing the 88 different frequencies (performed once per program execution only).
• Divide the distance reading by the above value to obtain table index.
• Obtain frequency from table (using index determined above).

 

This was the second mapping used (the first being linear as describe in section 4.3.1), and appears promising and is kept separately ('lrf_piano.c'),along with its own makefile ('makefilePiano'),  allowing for future use and/or modification.

4.4 DISTANCE TO MAGNITUDE MAPPING

Prior to moving to a PCM encoding, u-law encoding with an 8-bit sample size was used.  This restricted sample sizes to a range of -128 to 127.  Along with the problems encountered with scaling the samples by a value below 0.5 (see section 4.2), altering the magnitude with distance had a minimal effect on the output tone.  As discussed in section 4.2, the magnitude could be altered from 0.5 to approximately 2 which is a very small range.

Switching over to PCM allowed a far greater range.  In addition to eliminating the problem of multiplying samples by less than 0.5 (see section 4.2), the increase in sample size from 8 to 16 meant samples could range from -32168 to 3276.  Altering the magnitude of a wave is now far more effective.

Using PCM also creates a new problem.  Pure tones output appear to follow the Robinson-Datson "Curve of equal loudness for tones of different frequencies" (as shown in figure 4.5 below), even when the magnitude of the wave is not altered. 

The Robinson-Dadson curve represents the amplitude levels at which single sine tones of different frequencies sound equally loud.  For example, in order for a 100Hz tone and a 1000Hz tone to sound equally loud, the amplitude of the lower tone must be boosted by almost 40dB (Moore, 1995).  Although the goal was to alter magnitude inversely with distance, as shown in figure 4.5, the magnitude of a 3500Hz tone exceeds that of the 6000Hz tone (which represents the closest distance and should also contain the greatest magnitude), and is definitely perceived so.  This problem can be overcome if the maximum frequency is limited to about 4000Hz.  Frequencies increasing from 20Hz up to 4000Hz are perceived as louder.

4.4.1 INVERSE SQUARE LAW MAPPING

Using this mapping it was found that  the larger distances are greatly reduced to the point where they are not audible (when k is less than about 20).  For example, when k = 5, the magnitude corresponding to 10.00m is 0.05 causing the tone to be barely audible (when k = 0.01, it results in a non-audible tone regardless the frequency).  To overcome this problem, the value of k can be increased, however, this will also increase the magnitude of the tones corresponding to the smaller distances to a point where they are too loud.  Extremely loud tones may interfere with the (possibly important), environmental sounds the user hears (e.g. passing cars, people etc.).

To decrease the allowable range of magnitudes, several different exponent values (using the equation above with k=5), were tried as shown in table 4.4 and figure 4.6:
 

 

4.5 GENERATION OF THE SAMPLES

All values generated by the above cosine formula will range between 0 and 1 (this may not be the case when the constant 'c'is greater than 1). Since the audio port requires integer values, these samples are then converted to the appropriate integer values. The functions provided in the file 'audio_hdr.h' will
convert the above samples to 8, 16 or 32 bit values.

For each sample the cosine function (cos) is required.  In terms of processor time, the cosine function (available in math.h of the standard C library), is very expensive, especially when it must be called several thousand times to create the samples of a single cosine wave.  In an attempt to minimise the time required by the cosine function, a cosine look-up-table was created.  Prior to initiating distance readings, the cosine values for 100 radian values (increasing by 0.01from 0 to 1.0), were determined and stored allowing them to be 'looked-up' instead of using the cosine function.  When the cosine of a number is required, the table is searched for the corresponding cosine value.  If found, it is returned, otherwise, the average of the value below and above are returned.

Quite surprisingly, it actually took longer to use the cosine look-up-table, as opposed to using the cosine function (see table 4.5).  The cosine function may have its own look-up-table which is more efficient than the within implementation.
 

4.5.1 OUTPUT OF A CONTINUOUS TONE

The LRF was set to take 8 readings/second (0.125s/reading).  The time required for the program to obtain a reading, determine the frequency and generate and output the samples corresponding to this frequency is 0.118s.  To allow for a continuous tone the above process must be accelerated.  This prompted creation of  a wave-look-up table (see section 4.6 for a complete explanation), allowing the samples representing a pre-defined number of frequencies to be stored in a table and used when needed.  With the wave-look-up table in place, the time between the output of  successive tones is greatly reduced allowing the perception of a continuous tone.  Upon obtaining a distance reading from the LRF, the samples corresponding to it are looked up and continuously output until the next reading is available from the LRF.

Although a definite change in the pitch of the continuous tone can be perceived with changing distance, the output of the this tone can become annoying to the listener after several minutes.  In addition, during the output of frequencies in the range of 3000Hz - 3500Hz, a 'clicking' sound could be heard in addition to the tone, coming from the background.  This click could be caused from non-matching phase relations of the output tones.  In an attempt to eliminate this problem, outputting one period only for each wave ensuring correct phase relations was tried. However,  there was no sound at all to be heard when this single period was output.  (It was later found outputting a single period of a tone prevents the tone from being heard).

Outputting a continuous tone was decided against.  In addition to the 'click' problem described above, after several minutes, the sound was bothersome to the listener.

This version was also kept separately (lrf_main2.c), along with its makefile (makefile2).

4.6 REAL-TIME SAMPLE GENERATION vs. WAVE-TABLE LOOK-UP

Using real-time generation allows the mapping of each distance to a unique frequency cosine wave, whereas when using the table of samples, certain distances will respond to the same frequency output.  However, the difference between the frequencies decreases as the distances increase, (using the present logarithmic mapping), and since humans require a minimum change of frequency between two pure tones (about 2-5 Hz), to perceive them as different (Hollander, 1994), the real time generation is not essential.

Real-time generation is approximated using the look-up-table by increasing the number of frequencies stored in the table.  However, increasing the number of frequencies stored causes an increase in both  the memory space required,  as well as the time to initialise the table.   A look-up-table containing samples for 145 cosine waves was used.  This table contains waves for distances ranging from MIN_DISTANCE (0.5m) increasing by 0.10m until MAX_DISTANCE (15m).  The increment of 0.10m represents the accuracy of the LRF (see section 2.3.1.7).

With the present size table, it takes approximately 8 seconds to initialise. Nevertheless,  time is saved  using the look-up-table. In particular,  it takes approximately 0.091 seconds to obtain a reading, determine and output the corresponding samples whereas using real-time sample generation takes approximately 0.118 seconds (see section 3.9.3 for the time determination).  This decrease in operation time allows for an increase in the number of readings taken per second (shots/reading was changed from 150 to 120 as a result).

4.7 DEPTH DISCONTINUITY (DC)

Distance changes greater than  1.50m will result in a depth discontinuity.  It was found that using a value greater than this will prevent several instances where there is a depth discontinuity from being conveyed to the user. For example, by measuring the distance between several doorways and the wall beyond, there were  several instances where the depth discontinuity is quite small (e.g. less than 2.0m), as shown in table 4.6 below:
 

4.7.1 SUMMATION OF SINE WAVES

To achieve this summation effect with sampled waveforms, the following formula is used for each sample:

Using the above formula, several combinations of cosine waves were summed with the resulting wave used as a depth discontinuity signal.  Initially, summing non-harmonic waves was yielding a sound not much different from the pure tones used.

The next method used was approximating a square wave (whose spectrum contains an infinite number of odd harmonics of the fundamental whose amplitudes decrease in proportion to the increasing harmonic number).  Although an infinite  number of samples could not be used, using about seven allows a square wave to be approximated.  The following formula was used:

Using a square wave did not produce a change over the previous signal containing non-related harmonics.

4.7.2 TRIANGULAR AND PULSE SIGNALS

• Assign a value to the first sample (a smaller value, for example -32768 or 0)
• The remaining N-1 samples are assigned the same value (much larger than the first, for example 32767).
• The last sample is assigned the same value as the first sample.

To create a single triangular signal, several different methods were used, as described  below:

4.7.2.1 PROBABILITY:  THE TRIANGULAR DISTRIBUTION:

4.7.2.2 TRIANGULAR FORMULA:

c = constant
t = Sample rate period * number of samples  (duration of the signal)
n = ranges from 0 increasing by 1, to 15 the total number of samples

To create the second half of the wave, n was then decreased until it became zero.
The following algorithm was used to implement the above function:

• Determine the wave duration by multiplying the total  number of samples by the Sampling Period (time taken to take one sample).
• Determine the value of the constant 'c' by assigning an arbitrary value to sample[n] (representing the peak of the triangle) and solving the equation for 'c'.
• Using the value for 'c' obtained above, solve the equation for all samples ranging from 0 to 15 the total number of samples.

NOTE:
While experimenting with the triangular and pulse signals, using the u-law encoding, I found
that sending less than nine samples to the audio port resulted in a 'click' being output regardless of the value of the samples.  Sending over nine samples resulted in no sound heard.  I could not understand this effect as the duration of nine samples is far less than the minimum duration required to perceive a sound.
When sending data to any port using the write system call, the number of bytes to write must also be specified.  It took several hours to realise I was sending nine bytes and specifying 4000 bytes were being sent!
 
The pulse and triangular signals were clearly distinguishable from any of the pure tones or pure tone summations used previously. Yet, they both sounded the same, having very short durations and weak signals, making them barely audible.  In order to overcome this problem, the signal was output continuously for a set number of times.  By outputting the pulse, triangular signals or pure tones in loop a certain number of times, gives the effect of a continuous signal, with no breaks in between. As table 4.8 below shows, both signals took about the same amount of time to produce and output.
 

For both triangular and pulse signals, generating and outputting less than about 20 samples, results in a non-audible signal. Furthermore, as the number of samples are increased beyond 200, the signal begins to sound like noise.   Since both signals sound the same, currently the pulse signal is used to convey a depth discontinuity since it is easier to implement.

4.7.3 USING AN AVERAGE OF THE N PREVIOUS DISTANCES

• Obtain distance reading.
• Calculate average of the previous N readings.
• Compare the current distance reading with average of previous N readings.
• Replace oldest distance of previous readings with the current reading.
• Determine how many previous readings exist.

Instead of limiting the possibility of errors, it was found that  an average actually created errors in several circumstances, when a depth discontinuity, although existing,  was not detected.  For example, consider N = 5 and the following five previous distances:

with the current distance reading of 2.10m.  The average of the above five readings is 0.784.  Using the current distance difference required for a depth discontinuity (1.5m), there is clearly a depth discontinuity between the last of the readings (0.50m) and the current reading (2.1m). However, using the average of the above five readings (0.78m), does not result in a depth discontinuity
(2.10m - 0.78m = 1.32m).
Regardless of the value used to compare depth discontinuities, examples of the foregoing  can be easily found.  As a result, averaging the previous N readings was not used.  The function calcAverage() (see 'lrf_audiov2.c'), was used to calculate the previous readings.  Although it is presently not used, it is included in the event of future applications.

5. INTERNET SITES VISITED

5.1 RELATED RESEARCH (PAST AND PRESENT)

1.  http://www.psyc.nott.ac.uk/bmru/history.html

History of the Blind Mobility research Unit from 1965-1995 at Nottingham University.  Describes the development of Electronic Travel Aids, mobility for the blind in general and the involvement of the above research unit in the field.

2.  http://www.ul.cs.cmu.edu/books/mobility_aids

ELECTRONIC TRAVEL AIDS: NEW DIRECTIONS FOR RESEARCH.  A 124 page publication from the Working Group on Mobility Aids for the Visually Impaired and Blind. Committee on Vision.  National Research Council.  The following topics are covered:

• The Demography of Blind and Visually Impaired Pedestrians
• Assessment of Mobility
• Perceptual, Cognitive, and Environmental Factors
• Sensory Enhancement and Substitution
• The Technology of Electronic Travel Aids

 "Rank Prize Fund Meeting:  Technology to Assist the blind and Visually Impaired".  A conference held in the UK March 25-28  1996 of the There are no complete papers available however, abstracts describing several topics including navigation and mobility for the blind, are available.

5.1.1 OBSTACLE DETECTORS

1.  http://bme01.engr.latech.edu/cdrom/texts/321.html

DYNAMIC ULTRASONIC RANGING SYSTEM (~1993/94):

• A prototype built around a PVC tube resembling a long cane.
• Four computer controlled ultrasonic sensors provide real-time information on distance and height within 5m of the user.

5.1.2 ORIENTATION AND NAVIGATIONAL AIDS

1.  http://simsrv.cs.uni-magdeburg.de/~mobic/tideful.html

MoBIC (MOBILITY OF BLIND AND ELDERLY PEOPLE INTERACTING WITH COMPUTERS) (1995):

• Does not provide obstacle detection.
• Orientation and navigation aid through the use of GPS.
• Consists of two parts:

• helps a user plan a journey (obtains map information, bus/train schedules etc.).
• provides data specific to the blind traveller.

• Guides the user towards their destination using GPS and an electronic compass
• User can ask certain questions, for example "where am I", through a hand held keyboard.

PAM-AID (PAERSONAL ADAPTIVE MOBILITY FOR FRAIL AND ELDERLY BLIND PEOPLE); TELEMATICS APPLICATION PROGRAM (JAN. 97 - PRESENT):

• Uses technology from industrial automation.
• The objective of PAM-AID is to allow users to retain their personal autonomy.
• Supports three modes of operation:

• Issues warnings when an obstacle is detected and intervenes if there is severe danger.

• the system navigates unsupervised avoiding obstacles.

• system makes small adjustments to minimise the risk of a collision.

University of Hertfordshire Sensory Disabilities Research unit.  Includes information regarding the following projects:

• MoBIC (project member)
• GUIB (Graphical User Interface for the Blind)
• PAM-AID (project member)

Home page of Gerrad Lacey, a Computer Science graduate student at Trinity College Dublin, Dublin Ireland.  His research interests include "The application of mobile robot technology to the mobility of the frail visually impaired".  Several of his publications are available online including "Personal Adaptive Mobility Aid (Pam-AID) for the Infirm and Elderly Blind", "Autonomous Guide for the Blind' and "Evaluation of Robot Mobility Aid for the Elderly Blind".

5.   HTTP://www.engin.umich.edu/research/mrl/00MoRob_22.html

GUIDECANE (Guidance Device for the Blind)

• An array of ultrasonic obstacle-sensors attached to a wheeled handle or cane.
• The user selects a direction of travel by using a joystick on the cane and then pushes the cane  forward.  The user can then turn right or left moving the joystick in the respective location.
• When the sensors detect an obstacle, the best path around the obstacle is determined and the wheeels are turned accordingly.  Through the cane the user feels the change of direction and follows.  Once the obstacle is passed, the cane brings the user back to the desired direction of travel
• Uses an on-board 25MHz 486 PC.
• Weighs only 4kg.
• Contains 10 ultrasonic sensors capable of sensing any obstacles within 120o.

6.  HTTP://www.arkenstone.org/

STRIDER - A Talking Portable Navigation System: (by Arkenstone)

• Addresses the transportation needs of the visually impaired.
• Consists of a Global Positioning System (GPS) and Differential GPS receivers which plug into a notebook computer.
• The GPS receivers provide lthe system with the user's latitude and longitude positions which after processing, the location of the user is conveyed to the user through software speech.
• Currently a proto-type (packaged in a backpack), has been developped.

5.1.3 ENVIRONMENTAL AIDS

1.  http://ourworld.compuserve.com/homepages/Peter_Meijer

vOICe (1994):

• Converts images taken with a standard T.V. camera into sound patterns on a 1-1 image to sound mapping ensuring preservation of the visual information (Meijer, 1992).
• Each picture is sub-divided into 4096 pixels (64*64) and each pixel can be one of 16 shades of grey.
• Every column is then translated into a sound (top pixel is high pitched, bottom pixel is low pitched and intermediate pixel has intermediate pitch.  The shade of the pixel indicates the loudness), and then summed allowing them to be heard simultaneously.
• The visual image is scanned from left to right outputting the sound of each column.
• Pictures are taken and converted to sound every second.
• Real-time processing.

"APPLICATION OF MACHINE VISION TECHNOLOGY TO THE DEVELOPMENT OF AIDS FOR THE VISUALLY IMPAIRED":  Dublin University, School of Electrical Engineering.

• Combination of shape, colour and motion is used to provide an auditory description of an image obtained through a camera linked to the Prolog environment.

3.  http://ourworld.compuserve.com/homepages/Vuphonics

VUPHONICS (86 - PRESENT):

• Sound (coded phonetics, multiple voices), resembling a language (English of European), which the user recognises and assigns meaning to, leading to a mental image, are output.
• Factual information (e.g. colour), is also conveyed by sound.
• The user builds up the layout of a scene on a rectangular grid where the resolution varies (resolution increases near areas of attention).
• Images to be conveyed are stored as the sections of a grid and sounds representing the overall light characteristics of and relative distances of features in the sections of images are output in a structured manner.

5.1.4 VIRTUAL SOUND:

Master's thesis of Ari J. Hollander exploring the potential to perform shape recognition of virtual auditory objects. Contains a very good summary of auditory perception and mentions many previously conducted experiments.

5.2 Commercial Products

1.  http://www.sonicvision.co.nz

Sonic Vision - a company of Leslie Kay (a pioneer in the Electronic Travel Aid field).  Site contains information and diagrams of the following three products:

BINAURAL SENSORY AID (SONICGUIDE) (1965): 

• Environmental Aid.
• Pair of glasses containing an ultrasound emitter and two receivers (left, right), whose signals are fedseparately to the corresponding ear allowing for object location to be determined through binaural hearing.
• Distance up to 18ft.,is indicated inversely with pitch (higher pitch = greater distance).
• Quality of the output sound can be used to identify object surface (clear sound = hard surface, fuzzy sound = rough surface).
• Does not interfere with normal hearing.
• Can be used to teach perceptual development to blind infants and children.
• Does not detect drop-offs, curbs.

• Sonicguide^TM mounted on a head band, with an added high resolution monaural sensor in order to model the dual function of the eye (fovial and peripheral fields of view).
• Uses a micro-processor to generate high resolution location of objects.
• Location of multiple objects in a 90^o  field of view.

Tony Heyes' Perceptual Alternatives home page containing information about the Sonic Pathfinder.  Links to other sites related to blind mobility are also available.

SONIC PATHFINDER: 

• Environmental Aid with a range of about 8ft.
• Head mounted computer controlled sonar system containing three receivers.
• Notes of the musical scale are output progressing in tone as the object is approached.
• Object to the left or right of the user causes the tone to be heard in the respective ear whereas an object directly ahead outputs the tone to both ears.
• The presence of the nearest object is conveyed and without any object present, the user is informed about the proximity of the shore-line.

5.3 MIDI SITES

Complete MIDI tutorial including diagrams/pictures covering the following topics:

• MIDI vs. Digitised Audio
• MIDI Basics
• MIDI Messages
• Synthesiser Basics
• PC to MIDI Connection
• MS Windows Configuration

"Documents that explain MIDI concepts in non-technical terminology.  Also contains information that may be useful to musicians using MIDI based products."  Some of the topics covered include:

• Basic Tutorial
• Hardware Discussions
• Computer and Hardware Configuration/Setup/Troubleshooting
• Performance and Sequence Tips
• Useful Charts

Technical details of MIDI (including General MIDI, MIDI file format, MIDI sample dump standard, MIDI time code), as well as MIDI programming information.

5.4 MISCELLANEOUS SITES

ICAD (International Conference on Auditory Displays).  Contains conference proceedings for 1992/94/96, links to journals related to auditory research and perception.

2.  http://sound.media.mit.edu/AUDITORY/asamtgs

Abstracts from the Acoustical Society of America conferences (124-130th meetings).  Many topics related to auditory perception, auditory scene analysis psychoacoustics etc.

3.  http://www.rnib.org.uk

Home page of the Royal National Institute for the Blind (UK) containing general information on blindness and the needs of the blind.  Contains many links to other sites.
 
 4.  http://www.psych.not.ac.uk/bmru

Nottingham University, Department of Psychology, Blind Mobility research Unit.  Includes information describing their current:

NOMAD: Purpose of the project is to help visually impaired students find their way on campus
                through the use of an audio and tactile graphics map of the campus.

5.  http://www.robots.ox.ac.uk/~david/home.html

Home page of David Lee (member of the Robotics Research Group.).  Several papers related to the mobility of the blind are available including "Object recognition with FM Sonar".

5.4.1 LINKS TO OTHER RELATED SITES

New York Institute for Special Education.  Contains links to blindness related sites.

2.  http://www.cs.yorku.ca/~billk

Summary of some of the work conducted in the field of Electronic Travel Aids.  Includes a description of many products (both past and present).  Some of the information on this site appears in this report.

3.  http://datura.cerl.uiuc/netstuff/sigSoundLinks.html

Many links to sites related to the following:

• Auditory Research
• Computer Music
• Sonification and Virtual Environment
• Speech
• DSP
• Hyper Audio

"Links to 3D and spatial sound related resources on the web.  Reference, research, educational tools, people, commercial systems and products.  Links to other 3D related and general audio lists and patent resources".

5.5 NEWSGROUPS

comp.music.research
comp.music.dsp
comp.music.midi
comp.sys.sun.hardware
comp.sys.sun.misc
comp.lang.c

6. SUGGESTIONS AND CONCLUSIONS

6.1 IDEAS FOR FUTURE CONTINUATION OF THE PROJECT

6.1.1 USE OF MIDI TO GENERATE SOUND

6.1.2 USE OF A MACINTOSH COMPUTER OVER  THE SUN WORKSTATION

6.1.3 ELIMINATING THE EFFECTS OF THE ?CURVE OF EQUAL LOUDNESS?

6.1.4 EXPERIMENTATION

6.1.5 NEW 'POWER' AND 'DATA OUT' CONNECTORS REQUIRED

6.2 CONCLUSIONS

6.3 Acknowledgements:

7. REFERENCES

7.1 BIBLIOGRAPHY

Arias C.  C. A Curet.  H. F Moyano.  S. Joekes and N.  Blanch.  (1993).  ?Echolocation:  A study of
     Auditory Functioning in Blind and Sighted Subjects.? Journal of Visual Impairment and Blindness,
     87, 73-77.

Blasch, B B. G. Lang and N. Griffin-Shirley.  (1989).  "Results of a National Survey of Electronic
     Travel Aid Use." Journal of Visual Impairment and Blindness, 83, 449-453.

Bower, T. G. R. (1982). "The Use of SonicguideTM in Infancy." Journal of Visual Impairment and
     Blindness, 76, 91-100.

Brabyn, John.  (1982).  "Problems to be Overcome in High-Tech Devices for the Visually Impaired."
     Optometry and Vision Science, 69 (1), 42-45.

Brabyn, John.  (1985).  "Development in Electronic Aids for the Blind and Visually Impaired." IEEE
     Engineering in Medicine and Biology Magazine, December 1985, 33-37.

Bregman, A. S. (1990). Auditory Scene Analysis:  The Perceptual Organisation of Sound. MIT Press,
     Massachusetts,  U.S.A.

Buser, P and M. Imbert. (1992). Audition. MIT Press, Massachusetts, U.S.A.

Crispien, Kai and Klaus Fellbaum.  (1996).  "A 3-D Auditory Environment for Hierarchical Navigation
     in Non-visual Interaction."  Proceedings of ICAD 1996.

Dodds, G.  Allan.  (1983).  "The Sonic Pathfinder - an Objective Evaluation."  International Journal of
     Rehabilitation Research, 6 (3), 350-351.

Dodds, G.  Allen, J. D. Armstrong and C. A. Shingledecker.  (1981).  "The Nottingham Obstacle
     Detector:  Development and Evaluation." Journal of Visual Impairment and Blindness, 75, 203-209.

Dodge, Charles and Thomas A. Jerse. (1985). Computer Music, 2nd  Edition. Schrimer Books, New
     York, U.S.A.

Easton, D. Randolph.  (1992):  "Inherent Problems of Attempts to Apply Sonar and Vibrotactile
     Sensory Aid Technology to the Perceptual Needs of the Blind." Optometry and Vision Science 69
     (1), 3-14.

Fish, M. Raymond and Ronald C. Fish. (1976). "An Electronically Generated Audio Display for the
     Blind." Journal of Visual Impairment and Blindness, 70, 295-298.

Frisken-Gibson F. Sarah, Paul Bach-Y-Rita, Willis J. Tompkins and John G. Webster. (1987). "A
     64-Solenoid, Four Level Fingertip Search Display for the Blind."  IEEE Transactions on
     Biomedical Engineering, 34 (12), 963-965.

Frysinger, P. Steven and Gregory Kramer, (eds) (1996). Proceedings of the Third International
     Conference of Auditory Display, held November 4-6, 1996, Palo Alto California.

Gao, X. Robert and Li Chuan (1993).  "A Dynamic Ultrasonic Ranging System as a Mobility Aid for
     the Blind."  Louisiana Technical University.  Ruston (U.S.A).

Guarniero, Gerard.  (1977).  "Tactile Vision:  A Personal View." Journal of Visual Impairment and
     Blindness, 71, 125-131.

Heyes, D. Anthony. (1990). "The Sonic Pathfinder."  TechniTalk, Winter 1990.

Heyes, D. Anthony. (1984). "The Sonic Pathfinder:  A New Electronic Travel Aid." Journal of Visual
     Impairment and Blindness, 77, 200-202.

Hioki, Warren. (1990). Telecommunications. Prentice Hall, New Jersey, U.S.A.

Hollander, A. J.  (1994).  "An Exploration of Virtual Auditory Shape Perception."  Ph.D.thesis,
     human Interface Technology Lab, University of Washington, U.S.A.

Kao, Gordon, Penny Probert and David Lee. (1996) "Object Recognition with FM Sonar:  An Assistive
     Device for Blind and Visually Impaired People." American Association for Artificial Intelligence.

Kay, Leslie.  (1980).  "The Sonicguide, Long Cane and Dog Guide:  Their Compatibility."  Journal of
     Visual Impairment and Blindness, 74, 277-280.

Kay, Leslie.  (1984).  "Acoustic Coupling to the Ears in Binaural Sensory Aids."  Journal of Visual
     Impairment and Blindness, 77, 12-16.

Kendall, S Gary.  (1991).  "Visualisation by Ear:  Auditory Imagery for Scientific Visualisation and
     Virtual Reality."  Computer Music Journal, 15, (4) 70-73.

Koren (Dr.).  (1995).  Course Notes:  Protocols & Computer Networks. Tel-Aviv
     University.  (http://www.rad.com/networks/1995/rs232/rs232.html)

Kramer, Gregory and Stuart Smith, (eds) (1994). Proceedings of the Second International Conference
     on  Auditory Display, ICAD '94, held November 7-9, 1994, Santa Fe, New Mexico.

Lacey, Gerrard and Kenneth Dawson-Howe. (1997). "Evaluation of Robot Mobility Aid for the Elderly
     Blind." Proceedings of the Fifth International Symposium on Intelligent Robotic Systems,
     Stockholm Sweden.

Lacey, Gerrad and Kenneth Dawson-Howe. (1997). "Personal Adaptive Mobility Aid (PAM-AID)
     for the Infirm and Elderly Blind." Lecture Notes in AI:  Assistive Technology & Artificial
     Intelligence.  Mittal Vibhu, Holly Yanco and John Aronis. (eds).

Loomis, M. Jack, Chick Hebert and Joseph G. Cicinelli.  (1990).  "Active Localisation of Virtual
     Sounds."  The Journal of the Acoustical Society of America, 88 (4), 1758-1764.

Loomis, M.  Jack, Roberta L. Klatzky, Reginald G. Golledge, Joseph G. Cicinelli, James W Pellegrino
     and Phyllis A.  Fry.  (1993).  "Nonvisual Navigation by Blind and Sighted:  Assessment of Path
     Integration Ability."  Journal of Experimental Psychology:  General, 122 (1), 73-91.

Loomis, M. Jack, Reginald G. Golledge, Robert L. Klatzky, Jon M. Speigle and Jerome Tietz.  (1994).
     "Personal Guidance System for the Visually Impaired." Department of Psychology, University of
     California.

McKennell, A. Bruce and E. Walker.  Blind and Partially Sighted Adults in Britain:  The RNIB Survey.
     Volume 1, Her Majesty?s Stationary Office, London.

Meijer P.B.L. (1992).  "An Experimental System for Auditory Image Representation.". IEEE
     Transactions on Biomedical Engineering,  39 (2), 112-121.

Miletic, G.  (1994). "Vibrotactile Perception:  Perspective Taking by Children who are Visually
     Impaired."  Journal of Visual Impairment and Blindness, 88, 550-551.

Miletic, G, B. Hughes and Paul Bach-y-Rita. (1988). "Vibrotactile Stimulation:  An Educational
     Program for Spatial Concept Development."  Journal of Visual Impairment and Blindness, 88,
     366-370.

Miller, L.  (1992).  "Diderot Reconsidered:  Visual Impairment and Auditory Compensation."  Journal
     of Visual Impairment and Blindness, 86, 206-210.

Mims M. Forrest. (1973). "Sensory Aids for Blind Persons." The New Outlook, November 1973, 67,
     407- 414.

Molly, D, T. McGowan, K. Clarke, C. McCorkell and P. F. Whelan. (1994). "Application of Machine
     Technology to the Development of Aids for the Visually Impaired."  Proceedings Machine Vision
     Applications, Architectures and Systems Integration.  Oct. 31 - Nov. 2 1994, Boston U.S.A.

Moore, C. J. Brian, (ed) (1995). Hearing. Academic Press, San Diego California, U.S.A.

Moravesik, J. Michael. (1987). Musical Sound:  An Introduction to the Physics of Computer Music.
     Solomon Press, U.S.A.

Moricca, S. Larry and Rex V. Slocum.  (1977).  "Pattern recognition on the Forehead:  An Electronic
     Scan System."  Journal of Visual Impairment and Blindness, 71, 164-167.

Paturzo, Anthony Bonaventura. (1984). Making Music with MicroProcessors. Tab Books, Philadelphia
     U.S.A.

Plomp, R. (1976). Aspects of Tone Sensation. Academic Press, New York, NY, U.S.A.

Pohlmann, C. Ken. (1989). Principles of Digital Audio, 2nd Edition.  Howard W. Sams & Company,
     U.S.A.

Pollack, Irwin and Mitchel Rose, (1967). "Effect of Head Movement on the Localisation of Sounds in
     the Equatorial Plane." Perception and Psychophysics

Rossing, D. Thomas. (1982). The Science of Sound. Addison-Wesley Publishing Company, U.S.A.

Rothstein, Joseph. (1995). MIDI:  A Comprehensive Introduction. Maddison Wesley Publishers, U.S.A.

Scadden, Lawrence. (1969). "A Tactual Substitute for Sight."  New Scientist. 41, 677-678.

Schustermann, J. Ronald. (1965). "Echo Detection Ability of the Blind:  Size-Distance Factors."
     Journal of Experimental Psychology. 70, 245-251.

Scione, M. W (1978). "Electronic Sensory Aids in a Concept Development Program for Congenitally
     Blind Young Adults."  Journal of Visual Impairment and Blindness, 72, 88.

Simon, Permina Ellen.  (1984).  "A Report on Electronic Travel Aid Users:  Three to Five Years
     Later." Journal of Visual Impairment and Blindness, 77, 478-450.

Sterlow, E R.  (1983).  "Use of the Binaural Sensory Aid by Young Children." Journal of Visual
     Impairment and Blindness, 76, 429-438.

Uslan, M. Mark, W. Robert Smith, Kenneth Schreibman and Douglas R. Maure. (1983). Journal of
     Visual Impairment and Blindness, 76, 71-75.

Wael, Hajjar and Guilhot J. Pierre.  (1988).  "Ultrasonic Travelling Aid System for the Blind."
     International Journal of Rehabilitation Research, 11 (1), 97-98.

Wadsworth, A. Jay.  (1984).  "Electronic Travel Aids for Children:  Advantages and Disadvantages."  A
     paper presented at the "Insight in Sight" Conference.  Vancouver B. C. October 1984.

Warren, H David and Edward R Sterlow.  (1984).  "Learning Spatial Dimensions with a Visual Sensory
     Aid:  Molyneaux revisited."   Perception, 13, 331-350.

Watkinson, John. (1994). An Introduction to Digital Audio. Focal Press, Oxford Great Britain.

Welsh, L. Richard and Bruce B. Blasch, (eds). (1980). Foundations of Orientation and Mobility.
     American Foundation for the Blind. New York, U.S.A.

White, W. Benjamin, Frank A. Saunders, Lawrence Scadden, Paul Bach-y-Rita and Carter C. Collins.
     (1970). Perception and Psychophysics. 7, 23.

Winsor, Phil and Gene DeLisa. (1991). Computer Music in C.  Blue Ridge Summit PA, U.S.A.

Zahorik, Pavel. (1996). Auditory Distance Perception: A Literature Review. Unpublished Manuscript:
     University of Wisconsin - Madison.

7.2 SOURCE CODE REFERENCES

1. Greg Reid's Lego Project - York University Department of Computer Science (1992):
2. James R. R. Service LRF Project - lrfd.c (1994)
3. Sun Microsystems Corporation - soundtool.c