DE3486044T2 - AUDIO-DIGITAL RECORDING AND PLAYBACK SYSTEM. - Google Patents

Description

Conventional recording of sound and playback of sound carriers are carried out using analog electronic systems. The sound waves of a sound source to be recorded are converted into electrical signals in a 1:1 ratio; the sound waves find their analogy in the electrical current generated by the microphone or the preamplifier circuit used in receivers, record players or tape recorders. When playing sound carriers, the electrical signal is amplified and used to drive loudspeakers that convert the electrical signal into sound waves, whereby an electromagnet and the loudspeaker membrane carry out movements.

Similarly, the output of conventional sound recording and playback systems consists of electrical signals in the form of signal waves that are either cut into a vinyl medium or transferred to magnetic particles on tapes. During playback, the signal waveforms are converted into sound waves as described above. The accuracy of the reproduced sound wave is therefore directly dependent on the quality of the metal or plastic disk or record or the tape itself. Both the creation of disk copies and tapes and their Reproduction devices result in a degradation of the quality of the reproduced analog signal. Noise due to contamination, wear and the inevitable background noise of the medium itself is therefore unavoidable in recording and reproduction with systems using conventional analog to analog recording and reproduction technology. Recent developments in the field of audio digital recording and reproduction systems demonstrate efforts to reduce or eliminate this noise problem. Exemplary of such developments are the types of systems and equipment disclosed in the following patents: Meyers et al, US 3,786,201, issued January 15, 1974; Borne et al, US 4,075,665, issued February 21, 1978; Yamamoto, US 4,141,039, issued February 20, 1979; Stockham, Jr. et al, US 4,328,580, issued May 4, 1982; Tsuchiya et al, US 4,348,699, issued September 7, 1982; Baldwin, US 4,352,129, issued September 28, 1982. These systems can generally be characterized by the fact that they take advantage of the high computing speed of digital electronic computers. The signal waves that represent sound in such digital recording and playback systems are repeatedly sampled to obtain a serial data stream that is converted into binary code and assigns a numerical value to each sample. This can be visualized as a continuous curve being divided into a large number of very small segments. This process is reversed during playback when each numerical value of each segment is converted into an output voltage. If this process is carried out quickly enough, the human ear can no longer recognize that the sound wave representing the sound has been "broken up" and reassembled. When sound is recorded in digital binary code in this way, it is, like music, just a series of numerical values embodied by magnetic tracks on the recording medium which, when read by a suitable electronic device, are either "on" or "off" without intermediate position. Such binary signals are immune to distortion, errors and wear over time. All sources that normally cause noise in analog devices have been eliminated, ie there is no tape noise, no tracking errors, no surface effects. The signal to noise ratio is limited only by the digital-to-analog conversion circuit itself and the output stages, but more by the sensitivity of the mechanical or magnetic analog-to-analog switching technology.

Of course, these systems also have several disadvantages. An example system currently in use for recording master tapes in the recording industry has excellent audio qualities, which come from the very fast sampling rate of 50 kHz and the resolution of the binary code into 16-bit "words" for each sample. The problem with this system is that each sample must be stored in mass storage for playback. The storage system must therefore store 4,320,000,000 bits of information in the correct order for a 45-minute recording. Storage systems with such capacities are large, expensive and generally not suitable as a consumer item.

Efforts to solve the problem of storage capacity have led to the approach of reducing the resolution of each sample (fewer bits per "word") and simultaneously lowering the sampling frequency (to 12 kHz). Such reductions have reduced the necessary storage space by more than a factor of 4. However, the resulting fidelity of the output signal is often below the level acceptable for hi-fi recordings.

Another attempt, particularly supported by telephone companies, uses the reduction in the number of bits described above and, in addition, a limitation of the bandwidth of the input signal to a range most used by the human voice (3 to 8 kHz). In this way, a reduction factor of over 12 can be achieved, again accompanied by a reduction in sound quality.

The most recent attempts to solve the storage and sound loss problem have used laser recording technology with extremely high digital storage capacities. This approach has been partially successful, as sufficient playing times have been achieved with the increased storage capacity. However, the manufacturing technology and equipment currently required to create a "laser-burned hole", "pit" or "black spot" in the storage medium is limited to "laser disks" or "laser fiches" in pure playback mode without the ability to record or erase and write at home.

The present invention therefore has the task The aim is to provide a system for hi-fi recording and playback that does not have the aforementioned disadvantages and associated problems.

Summary of the invention

The present invention, as defined in appended claim 1, is another approach to solving the storage and reproduction problems encountered in digital audio recording and playback systems as described herein. High fidelity can be achieved with limited computer memory space by providing a single electronic signal processing device that

1. converts analog data into digital data stream samples;

2. Selecting portions of the data stream samples to generate at least three data streams indicative of amplitude, frequency and waveform characteristics;

3. stores data samples which together describe a waveform and have a predetermined time duration, each of these data samples being compared with predetermined waveform parameters to select and retain only predetermined parts, the waveform data samples corresponding to the predetermined waveform and the real-time data and a new resulting waveform data code being generated from this comparison, the selected data from the data streams indicating the frequency and amplitude being then compared with the waveform data code, a further data code being generated which is proportional to the frequency and amplitude of the original analog signal, wherein the data stream determining the amplitude, the data code specifying frequency and amplitude, and the data code describing the waveform are sequentially recorded on a recording medium in order to subsequently play back the sound carrier by processing the sequentially recorded data.

Short description of the drawings

Fig. 1 is a schematic diagram of the digital recording and playback system of the present invention.

Fig. 2 is a pictorial representation of the analytical model of the operation of the data acquisition module of Fig. 1.

Fig. 3 contains a diagrammatic representation of the recorded waveform data.

Fig. 4 is a pictorial representation of a single unit of binary code as stored on the diskette and from which a reproduction is obtained in accordance with the operation of the present invention.

Fig. 5 is a diagrammatic representation of the layout of the electronic components used in the present invention.

Fig. 6 is a pictorial representation of an implemented example of a goods detection system.

Fig. 7 shows the analog output signal of the devices of Fig. 6.

Detailed description of the invention

The present invention includes a system for converting analog input signals, such as audio signals, into digital signals and then encoding them into structured data pieces for recording in compressed digital form and producing a digital record similar to the original digital input signal before conversion back to analog signal form.

In the broadest sense, therefore, the recording of the audio signals in digital form for subsequent playback is achieved by using a microcomputer recording system 2 which includes electronic components for converting an analog audio signal into at least three digital data streams: the first of the digital data streams is a wideband reference signal describing the amplitude of a predetermined range of audio frequencies; the second of the data streams is generated by filtering the analog audio signal to generate at least one data stream channel designated for a selected bandwidth of frequencies which are narrower than the bandwidth represented by the first data stream; the third data stream describes the selected frequency of the audio signal, dividing means for generating a sequential stream of selected data from each of the digital data streams, selecting means for selecting a predetermined portion of the selected digital data generated by the dividing means in each of the data streams; means for separately storing each selected digital data portion generated by the dividing means; means for comparing the reference signal data stream containing the amplitude data with the second data stream containing the frequency data to produce frequency spectrogram data describing the frequency and amplitude of the original audio signal; means for converting the selected data from the third data stream channel taken from the narrower bandwidth into data describing a histogram in which the amplitude is plotted against time for each bandwidth; means for comparing the histogram data with selected waveform parameters and for generating and storing addressable data describing the waveform of the original audio input signal; and means for sequentially assembling and storing the frequency spectrogram data and the amplitude reference data of the first data stream and the addressable waveform data for subsequent playback.

In the preferred embodiment of the invention, shown in Fig. 1, the input signal 3 is conditioned and amplified in the first section of the data acquisition module (DAM) 4. The DAM 4 is a multi-channel programmable microprocessor-based device that uses standard ICs to perform three functions:

1. Selecting samples with a frequency of 42 kHz, short-term storage, digitization and output of the wideband audio signal range (20 Hz to 20 kHz) of amplitude (DC voltage) every 0.01 sec. Thus, a digital "word" consisting of four to fourteen bits is generated one hundred times per second for assembly as part of a recording file on a diskette or disk.

2. Select, temporarily store, digitize and output an audio frequency spectrogram every 0.01 seconds from a 128-segment matrix of logical bandpass filters selected from 128 channels arranged logarithmically over the entire bandwidth used. The data set produced by this function can range from zero (no signal on any channel) to (n)* [7-bit identifier)+(7-bit scaler)+(2-bit pointer)], where (n) is the number of channels on which a signal is present.

3. Use as an input unit for a digital storage oscilloscope on which data strings of digitized amplitude versus time are displayed (histograms) corresponding to the range of the selected bandpass filters described in paragraph 2. This linked data set is generated every 0.01 sec., is the largest single data structure and contains a continuous listing of each active bandpass filter. The number of "words" in each data string is a function of the filter center frequency and requires not less than 4,000 selected sets for a 20 kHz channel or not less than five selected sets for a 20 Hz channel. This data set is not sent to the File assembler as in paragraphs 1 and 2 above, but loaded into a RAM memory 6 which can be accessed by the Waveform Analysis and Coding Module (WAC) 8.

The function of the Waveform Analysis and Encoding Module (WAC) 8 is that of a digital numerical processor section which is programmed to extract characteristic waveforms from the data set stored in RAM 6 by the DAM 4. The waveform data is reduced to a tabular form in which one period of each encoded waveform is assigned to a wave table which is preferably a digitized X-Y coordinate system consisting of 1,024 bytes on the X axis and an 8-bit "word" at each byte location, so that the Y axis is divided into 256 parts; 127 in the positive and 127 in the negative range. Therefore, a set of wave tables is generated for all active bandpass filter channels every 0.10 sec. In each cycle (0.01 sec.) a range from zero to 128 tables can be generated.

The WAC 8 uses one of several P.M.S. reductive analytical methods to find waveforms. The first is the "Fast Fourier Transform" (FFT) and the second is the "Fast Delta Hadamard Transform" (FDHT). The two methods can be briefly described as follows:

The FFT is based on the theorem that almost every periodic function f(x) of second order oscillations can be represented as a term with sine and cosine functions. In general, the equation is:

The algorithm for effectively solving this equation was first developed by Rabiner Gold in 1974 and by Oppenheim and Schafer in 1975.

The FDHT is used to analyze the spectral composition of the data set, where the spectrum Ψ has the form:

where Fi is the frequency and Ψi is the signal intensity. In the present application of this method, the digital output of the logic filter from paragraph 2 above is summed at each filter and added to the next output until all frequencies have been detected. In the last step, the total output is:

An approximation of the spectrum (Ψ') is obtained by matrix multiplication:

The algorithm for performing the FDHT was published in 1983 by E.E. Fenimore at the Los Alamos National Laboratory.

B-spline approximation algorithms can also be used to extract the characteristic waveforms become.

The last generated set of waveform tables is sent ten times per second to the Disk Record Assembler 10 (DRA, Fig. 1).

The Disk Record Assembler (DRA) 10 is a program module that receives as input every 0.10 seconds the addresses of the waveform tables from the WAC 8, just as it receives every 0.01 seconds the data sets for the frequency spectrograms described in paragraph 2 above directly from the DAM 4, as well as the digital "word" that characterizes the total broadband field strength. The waveform tables are held in a local memory buffer (LMB) 11 in the DRA 10 so that they can be changed or deleted every 0.10 seconds by a subroutine commonly called Waveform Catalog Maintenance (WCM) 12. Disk records (Fig. 4) can be of different lengths, but always have the same format: the first 14 bits are the field length specification 14, the next 7 bits are the frequency filter or channel detector 16, followed by a 2 bit pointer and its 7 bit scaler 20, then the 7 bit waveform table detector 22, the 7 bit simultaneous waveform table detector 24 (repeated if necessary), the 2 bit flag 26 (signals), then the next filter detector, and so on, until the last 14-bit "word" which identifies the wideband signal level. The format of the data stream is shown graphically in Fig. 4.

If a recording is prepared for storage, it is stored in a "local memory buffer" in the DRA 10. cycle time and can thus be compared with the next sequential recording. This allows the DRA 10 to use "tokens", specific reserved symbols to detect "repeats", "equals" and "zeros" in the current recording, which have been concatenated to save memory space.

The Waveform Catalog Maintenance subroutine 12 is programmed to check incoming waveform table "updates" against the already stored waveform tables and against each other. Since there are only 128 channels available for storing the DAM 4's amplitude histogram, comparing the WAC 8's waveform output with the DRA 10's stored waveform data determines what is redundant and deletes duplicates. The remaining incoming tables are either unique or slightly improved versions of shapes already stored. Waveforms that have new characteristics over existing tables are stored in place of the previous resident table. Unique or peculiar shapes are assigned to new tables. If an overload occurs due to full tables and a unique waveform arriving, it is stored in a "local buffer" for one cycle time, where it is checked for repetition. If it does indeed repeat in the next 0.10 second cycle, space is made for it in the waveform catalog by deleting one of the two waveforms that are most similar in the catalog. The algorithms for these estimates are based on standard statistical methods for measuring the similarity of two curves.

In the preferred example of the present invention, the storage medium is a 5 1/4" magnetic diskette, as is commonly used for digital magnetic storage and retrieval of data. These diskettes have a storage capacity of more than one megabyte (1 million bytes or 8 million bits), and in the near future this will probably be ten megabytes. A five megabyte diskette will be used for image data storage.

Assembled disk records from the DRA 10 are the inputs to the disk read/write module 28. In the write mode, records are written to the disk memory 30 in the form of the previously described data stream format as long as space is available. Considering that an average record consists of 20 bytes of data, the disk contains 240,000 records, each of which represents 0.01 second of real time. In addition, the entire waveform catalog 12 is written to the disk after the entire disk has been filled, excluding the 130 kBytes required for the waveform catalog itself.

In the retrieval or playback mode, the read/write module 28 first reads the waveform catalog 12 from the diskette or disk 30 into the RAM memory 6. The playback module 32 then accesses the waveform tables when it is called up for each recording. Every 0.01 seconds, a diskette or disk recording is serially read from the diskette or disk 30. read to preserve their relationship to the real time of the original audio source material.

The playback module 32 used in the present invention preferably includes digital oscillators 33 for generating the output signal 34 and "smoothing" filters 35 for eliminating the "steps" that occur in the digital representation of continuous analog functions. The logic of the playback module is based on the principle of "additive synthesis." Briefly, the theory of "additive synthesis" states that complex musical signals can be determined by adding together all of the voltage components of the signal at each individual frequency at each instant. Consequently, if the data reduction process retains all of the original voltage-over-time information in such a way that it can be precisely reassembled and is not time-shifted, the output signal will be equal to the original input signal at each specific frequency or "pitch." In the exemplified example of the invention, these "additive synthesis" conditions are at such a level of resolution that the human ear cannot distinguish most sources from the original. The playback module 32 then causes the oscillators to set the timbre of each oscillator at the frequencies specified by the disk recordings using the waveform reference data and the wideband amplitude reference data to set the voltage levels. Timing agreement is achieved in the system by establishing the 0.01 second cycle time upon which the system is based.

The most preferred implementation of the system uses Very Large Scale Integrated Circuit (VLSIs) technology to reduce the number of individual semiconductor chips, as shown in the schematic diagram shown in Figure 5, using many IC devices that are commercially available.

Referring to Fig. 2, the analytical model is graphically represented. The model has three reference axis dimensions to be measured: time, amplitude (DC voltage), and frequency. The time axis is divided into 0.01 second intervals. In understanding the operation of the present invention, it is important to recognize that the 0.01 second interval corresponds to the frequency at which incremental, acoustic "snapshots" of the audio signal are recorded. This increment was chosen because it is short enough that the human ear physiologically hears a sequence of 0.01 second interval changes as a continuous whole. The stream of acoustic "snapshots" is analogous to the stream of images in a movie.

The acoustic "snapshots" themselves contain in binary form the full range (20-20,000 Hz) of amplitude, a frequency spectrogram and references to the waveform table received from the DAM 4 (Fig. 1). The representation in the amplitude histograms in Fig. 2, such as (ah₆) 36, show the waveforms contained in the so-called "amplitude histograms" which are used as unprocessed data sets for writing the waveform tables. This will be discussed in more detail below. The recording of the entire wideband amplitude consists of reading (every 0.01 sec.) a continuous signal stream of 14 bit "words""written" by the wideband section, storing them temporarily and digitizing them at a rate of 42,000 "words" per sec. In other words, only one of every 420 "words" generated is stored. These chains of amplitude samples are used by the RAM buffer module 6 in "playing back" the digital oscillators at the output end of the system. Each amplitude read into each frequency channel is scaled to this reference level. Referring again to Fig. 2, the (BBR) "wideband reference record" 38 is a two-dimensional data array in which the first dimension indicates the value for the time within the 0.10 sec. time frame, which is expressed in hundredths of a second. (e.g. zero; 0.01; 0.02 sec.). The second dimension represents the binary representation of the DC voltage level or amplitude in each time increment. The voltage level is recorded with the accuracy of a 14-bit "word". Thus, there are 16,384 different values to represent the DC voltage range, which in the characteristic range of 0.5 volts to 5 volts is e.g. 100 dB. The absolute accuracy is therefore 4.95 divided by 16.384 or +/- 0.0003 volts DC.

It would be desirable to have such accuracy for recording to measure the DC voltage in each bandpass filter channel. However, in order to save storage space, it would be desirable to use as few bits as possible to represent the amplitude of the signal on each channel. To achieve these conflicting goals, the relative representation method is adopted. Each record of a frequency channel amplitude is a bit "word" called a divider value, thus 128 values are possible that record the signal of each channel relative to the wideband value. Thus, a channel with a DC voltage of 0.250 volts with a wideband value of 3.250 has a ratio value of 0.07692 considering the wideband signal. With a 7 bit scaling this is a "3" with a maximum value of 128. The second advantage of this approach is the increased processing speed achieved by the comparative nature of the modular arithmetic logic as opposed to the more time consuming logic of storing and encoding a 14 bit accurate "word" on each channel. So by using a 7-bit "word" instead of a 14-bit "word" 50% of storage space is saved.

Referring to Fig. 2, the frequency spectrogram Fs 10 is similar to the wideband amplitude recording except that the amplitude of the voltage in each of the 128 discrete, narrow bandwidths is stored every 0.01 sec. The 128 channels are selection, short-term storage and digitization circuits limited by the bandwidths they can "hear" due to their preselected bandpass filters. The distribution of the channels over the 20 to 20,000 Hz range can be controlled by the user via slide switches common to equalizers or by automatic signal searching. That is, the logic of the 128-channel range responds to the amplitude of the voltage in each of the 128 individual narrow bandwidths and "eigencenters." around the "live" bandwidths. This principle is the same as in signal-seeking radio receivers.

As shown in Fig. 2, there can be overlaps between channels, as shown by the hatched triangles on the frequency axis of the spectrogram. A signal in the overlap region logically indicates to the system that the channel range is not tuned to the incoming digital signal and can be used to set a corrective flag value that can be used by the auto-ranging circuit to move to the next "snapshot" to read the center of the channel.

The amplitude histograms, e.g. ah6 in Fig. 2, are generated whenever a channel is "active." These histograms are binary point-by-point plotted on a continuous real-time basis, with amplitude plotted against time. They are not "snapshots" taken every 0.01 second. The actual time required to plot a histogram varies with the audio frequency of the channel. It is generally accepted that the higher the frequency, the more data points are required to "feed" the waveform analyzer and encoder. Of course, the upper time limit for this process is 0.10 seconds, otherwise the synchronization of the entire system would be compromised. The purpose of the amplitude histograms is to provide the "unprocessed" data for FFT or FDHT routines that the WAC 8 uses. In order to use the FFT to calculate a sequence of XY coordinates as a periodic In order to characterize a curve function, at least two complete cycles of the periodic function must be collected. Because of delays in the electrical recording circuit, often referred to as "deposition time" noise, in many cases more than two cycles must be collected for analysis.

Referring to Fig. 3, the wave table catalog information is graphically presented in the form most suitable for the system. Once the waveform analyzer and encoder 8 (Fig. 1) has "found" a waveform in an amplitude histogram 36 (Fig. 2, ah₆), the waveform data is plotted for one waveform period in an X-Y coordinate system as graphically represented in the wave table wt₁ in Fig. 3. The amplitude of the wave is plotted in the Y direction using 1,020 8-bit binary "words" which allow for an accuracy of 127 steps above and below the X axis. The X axis itself is an arbitrary division of the waveform period into 1,020 increments. In the wave table, 4 bytes are reserved for information about the status of the table within the catalog of 128 tables. This is necessary because wave table position notes are created with each 0.01 second acoustic "snapshot" and can be changed as the recording progresses and more or better information becomes available. Preferably, all overwrites of the wave table notes should be made at the end of each complete recording session with one pass through the disk recordings.

Referring to Fig. 4, the bit scheme of a typical acoustic "snapshot" is graphically represented. An average diskette contains 240,000 of these acoustic "snapshots". The first binary word 14 comprises 14 bits and indicates in binary the number of the total number of bytes that follow in the acoustic "snapshot". This indication of the field length is necessary in order to be able to control or regulate the data flow in the system during playback. The playback module 32 must be informed by the controlling or regulating software how much data must be taken into the buffer memory for a real-time output of 0.10 seconds in length. The following 7 bit "word" 16 indicates to the playback module the first of the frequency detectors that are contained in the acoustic "snapshot" and which are followed by their 2 bit identifiers for frequency shifting, e.g. B. to indicate whether or not a shift is necessary and in which direction it should be made. The third 7-bit "word" 20 is a binary number (from the range 1-128) that sets the relative amplitude (voltage level) for the previously specified frequency output. The fourth 7-bit "word" 22 is a binary number (from the range 1-128) that tells the playback module 32 where the waveform addressed at the previously specified frequency can be found in the waveform table. The fifth and sixth 7-bit "words" 24 also indicate waveform table references that are used in the initial frequency setting.

In operation, the playback module 32 reads the acoustic "snapshots" (disk or disk recordings) and then passes them on to "load" a digital oscillator circuit with the values that it by cues and also by reading directly from the acoustic "snapshot". For example, in the case of the recording shown in Fig. 4, four frequencies are required. Each of these has a known number of oscillation frequencies which are the same as the channel bands. These frequencies are assigned to specific digital oscillators. The amount of energy that must be expended to operate the oscillator is determined by the relationship between the scaler and the wideband reference signal which in turn is identified by the 14-bit "word" at the end of the acoustic "snapshot". The waveform table cues are connected to the frequency oscillators in exactly the order in which they appear in the acoustic "snapshot". Each oscillator receives the characteristic tone which is identified by the sum of the waveforms applicable to it from the playback module 32. The frequency at which a wavetable can be played per sec. read by the playback module 32 for an oscillator is a function of the frequency of the oscillator, e.g. a 440 Hz oscillator reads a wave table 440 times per second.

Referring to Fig. 5, there is shown a schematic layout of the components and their connecting lines for a preferred embodiment of the invention. Comparing the requirements for the various functions shown in Fig. 1 with the capabilities of various electronic components available on the market, the present invention can be implemented after selecting components that meet the required requirements. For example, the broadband digitizer 40, which is used in the "data acquisition module" (DAM) 4 may be selected from commercially available high speed analog to digital converters such as those manufactured by Hewlett Packard, Datel, Inc., Intel or RC Electronics. The 128 channel array 42 in the DAM 4 for performing analog to digital conversions may also be obtained from Hewlett Packard, with specific pass-through characteristics for each individual channel being considered. 16 channel devices are currently available, so eight such devices are required. As already mentioned in the figures and this description, devices such as the RAM 6 should have a capacity of up to 500 kBytes, the Read Only Memory (ROM) 44 320 kBytes and the Central Processing Unit (CPU) 46 shown in Fig. 5 should preferably have a 16 byte/8 MHz capacity with several additional 80 byte numerical processors. The illustrated disk drive unit 48 may also be replaced by other suitable digital read/write devices such as magnetic bubble memory packs, linear magnetic tape drives, laser-based disks or fiches. The control panel 45 may include control of tone, gain, range, track selection and additional features, or it may be as simple as "on/off", "record/play". Signal input 47 and output 49 are via conventional RCA clamp jacks.

The preferred embodiment of the present invention has been described with emphasis in places on the required functional features of the components used in the system logic employed. It should be apparent that a variety of different Components in the general system described can be used to achieve similar or the same results. For example, different sample sampling rates or the like can be advantageously achieved with other components without departing from the spirit of the invention described. Of course, the recording and playback functions can be integrated or separated, and of course it may be possible to use the disclosed recording format with a computer or a data transmission device, such as a modem, to transmit the recorded data to a remote playback device.

In addition, digital information can be sent to a remote location via radio signals to supply a local storage unit with the required wave table information, and the information in the digital records of the recording can then be transmitted to the remote location for playback. This can be achieved by relatively narrow bandwidth transmissions. The audio signal can then be reproduced by the playback unit at a remote location based on the information sent via radio signals. Again, this system has the advantage of achieving a high signal-to-noise ratio, which is the case with any digital communication system.

The preferred embodiment of the present invention has been described using technical terms for recording and reproducing sounds. In addition to Using the described system as an audio playback system, the output signal can also be used to stimulate the listener's nerves, as achieved by hearing aids or artificial ears. The software would not need to be modified for this application.

The same concepts underlying the present invention can equally well be applied to medical acoustics, EEG or ECG analog signals. Waveform tables found during a basic medical examination using these non-surgical diagnostic methods can be used to digitize and condense the analog signals encountered during stress, post-operative or post-treatment diagnostics. The medical input could also be via an ultrasound probe, and the present invention could also store ultrasound images and examine them for density or other observable sound wave properties.

The embodiment of the invention can also be an acoustic alarm system. The data streams of the present invention would thus be indicative of security sound information. The RAM memory would provide data for comparison between potential intrusion sound patterns and previously collected sound recordings. This would eliminate false alarms from the alarm system. Storage disks could be changed according to the usage factor of the secured area. For example, a "weekend" disk would contain completely different information than a "night shift" disk.

The input device could be a vibration meter acoustically coupled to a piece of machinery, such as an electric dynamo, a mill or press in a manufacturing area, or other equipment containing bearings subjected to heavy or cyclic loading. The analog input contains numerous frequencies, amplitudes and waveform information indicative of the nature of those bearings. This information can be analyzed by the microcomputer system of the present invention and used to detect otherwise gradual changes in these signal parameters to predict imminent bearing failure or the like.

Fig. 6 shows yet another possible use of the microcomputer system described above and embodied in the invention. In this example, a system is included which can be used to collect and analyze inventory levels. The input to the data collection module of Fig. 1 is via a hand-held area finder 54. Preferably, the device 54 also includes a holographic type bar code reader. The device 54 is connected to a microcomputer system 56 (preferably a type provided with a battery power supply unit to make it completely portable) which includes the functional system shown in Fig. 1.

Goods G are placed on shelves for registration. The shelves should have a bar code B at at least one end of each usable shelf space. The goods should, if there are several goods should be loaded from the back of the shelf with small spaces between the goods for ease of location by the device 54. The bar codes are commodity codes or other specific markings describing the goods stacked on the bar coded shelf. Alternatively, a keyboard may be used to enter this information before each shelf is scanned by the device 54. In operation, the hand-held microcomputer goes to its home position and the disk recording assembler module and other control codes are entered. The operator would stand in front of the goods on the shelf and, if necessary, enter the commodity description (commodity code, shelf number, etc.) via the keyboard. If a bar code reader is incorporated in the device 54, this step is not necessary. The device 54 is passed past the goods stored on the shelf and moved the entire length of the shelf. A typical example of the resulting analog signal is shown in Fig. 7. This signal corresponds to the depth of space along the shelf filled with goods, which is a function of time determined by the echo of the rangefinder. The time interval in this case corresponds to the time between finding the bar code on the shelf side wall (if present) and finding the second shelf side wall or the time between two shelf side walls.

Again, the signal contains frequency, amplitude and waveform characteristics that are manipulated, altered and compressed to produce a digital disk recording as described above.

This digital recording can later be sent via a conventional modem 60 to a central goods registration point, where the data can later be displayed in graphic or tabular form.

An example of a non-acoustic output using the data compression and recording of the present invention would be to use it to record, compress and analyze seismographic data. In this case, the result would be a "doctored" graph with properties over a specific size and less noise or reverberation beyond a certain order of harmonic series.

In addition, any sensor that outputs an analog waveform type signal can use the present invention to compress and record data. Examples of such sensors would be pressure transducers, flow meters, photodiodes, microwave receivers (radio and radar frequency), photocells, piezoelectric devices, charge coupled devices, and spark counters. Likewise, digital data representative of waveform data can be compressed in accordance with this invention by first converting the digital data into analog signals which can then be processed by the system described herein using the methods disclosed.

The process methods used in the system described for division, analysis, coding and recording and for subsequent decoding and playback allow the system to achieve a storage density that is up to 300 times higher than the storage density existing systems. The scope of the invention should therefore only be limited by the state of the art, as set out in the following claims.

Claims (5)

1. A microcomputer recording system for recording analog signals in digital data form, characterized by conversion means (4) for converting an analog signal into at least three digital data streams, the first of said digital data streams being a reference signal representative of the amplitude in a relatively wide frequency range, and the second of said data streams being generated by filtering the analog signal to generate a data stream representative of a plurality of discrete frequencies encompassed by the bandwidth of the first data stream, and the third of said data streams being a reference signal representative of the amplitude of the analog signal for each of the pluralities of discrete frequencies; sampling means (4) for generating a sequence stream of digital data samples in each of said digital data streams; Selection means (4) for selecting a predetermined portion of the digital data samples generated by the sampling means in each digital data stream; means (10) for separately storing each of said predetermined portions of the digital data samples; means (6,46) for comparing said predetermined portion of the first reference data stream, the amplitude data with the predetermined portion of the second reference data stream containing frequency data to generate frequency spectrogram data indicative of frequency and amplitude of the original analog signal; means (6,46) for converting the predetermined portion of the digital data samples selected from the third data stream into data representative of a time versus amplitude histogram for each discrete frequency; means (6,46) for supplying selected waveform parameters; means (6,46) for comparing the histogram data with the selected waveform parameters and for generating addressable waveform data representative of the waveform of the original analog signal; and Means (46) for sequentially assembling and storing the frequency spectrogram data and the amplitude reference data of the first data stream and the addressable waveform data for subsequent use. 2. A microcomputer recording system according to claim 1, comprising playback means (32) for converting the digital data streams to analog data for subsequent playback, characterized by addressing means (37) for selecting the stored addressable waveform data representative of the original analog signal; and oscillator means (32,33) for generating an analog signal responsive to the addressable waveform data, to the frequency spectrogram data and responsive to the amplitude reference data to produce a resulting analog signal suitable for use in a playback device. 3. A system according to claim 1 or 2, characterized by the means (8,10,12) providing waveform analysis means including means for generating waveform tables using B-spline types of cubic and cube spline transformations, respectively. 4. A method of generating an analog signal from digital data provided by a microcomputer according to any one of the preceding claims, characterized by digital data including assembled waveform tables generated from histograms of characteristic waveforms of the original analog signal, from frequency spectrograms and from full spectrum digital amplitude signals and reproducing in the following steps: - addressing waveform, frequency and amplitude data provided by the histograms, spectrograms and amplitude signals, and - converting the waveform, frequency and amplitude data from digital to analog form. 5. A method according to claim 4, and the step of converting the digital data back to an original analog signal that can be played in an audio playback device, characterized by the steps of: - converting an original analog audio signal into at least three digital data streams, the first of said digital data streams being a reference signal representative of the sum of the amplitudes of a preselected series of audio frequencies in a relatively wide bandwidth, and the second of said data streams being generated by filtering the original analog signal to generate a plurality of frequency channels indicative of a number of discrete frequencies encompassed by the bandwidth of the first data stream; and the third of said data streams consisting of a number of reference signals representative of amplitude signals from frequency channels of the second data stream; - selecting a predetermined portion from each of said digital data samples of each digital data stream; - storing the selected, predetermined portions of the digital data samples; - comparing the predetermined portion of the first digital data stream containing amplitude data with the data stream containing frequency data to produce frequency spectrum data representative of the frequency and amplitude of the original analog signal; - converting the predetermined portion of the data samples selected from the third data stream into data representative of a time versus amplitude histogram for each discrete frequency; - Comparing the histogram data with the selected waveform parameters and generating addressable data representative of the original analog signal; - subsequently assembling and storing the frequency spectrogram data and the amplitude reference data of the first data stream and the addressable waveform data for subsequent use; - addressing the stored addressable waveform data, the frequency spectrogram data and the amplitude reference data representative of the original audio signal to generate a resulting signal therefrom; and - converting the resulting signal from digital to analog form.