What is a spectrogram?

                Sound – invisible, always in motion and intangible. Our ears perceive sound as a series of waves with changes in amplitude, frequency and harmonics moving through the air. The intangible properties and constant motion of sound are the exact reason why it is so hard to interpret sound patterns. Spectrograms change all of that. Spectrograms give us the power to take a sound or series of sounds, plot them in real time and make an interpretable data formation out of them.

                Spectrograms, also known as sonograms, are computer programs that convert the varying intensity of an acoustic signal into a visual representation that is plotted on a graph according to its intensity. In most traditional spectrograms a grayscale pallet has been used, however as times and technology has changed so too has the spectrogram. In the grayscale format the degree of darkness at any given point is directly proportional to its amplitude (sound energy) at that time and frequency. Most modern spectrogram software programs are created with a color pallet that displays shades and hues that are representative of the sound energy. The intensity of the sound energy varies with the colors in the pallet within the spectrogram program and thusly creates a more easily identifiable spectrum of data for the interpreter. The color intensity representations vary from one spectrogram software program to another so be sure to completely read the manufacturers instructions on how to interpret their color pallet. This “picture of sounds” is an incredibly useful tool to a paranormal investigator, especially when it comes to EVP work. 

10.2 – This is a basic illustration of the qualities of a spectrogram

                Spectrograms use a technique called Fourier analysis to convert the sounds into pictures. More precisely, a Fourier transform is applied to the electronic analog representation of an acoustic wave where the spectrogram then derives the translation of the frequencies and amplitudes of its component simplex waves into an interpretable data form. When this process is performed it renders a picture of sound that we can then interpret into meaningful data. 

                Depending on the depth of detail programmed into the analytical properties of the Fourier analysis window of the Spectrogram there can be a variety of different levels of resolution and detail that can be observed from one spectrogram to the next. Most spectrographic programs use a short-analysis wide-band window where the adjacent harmonic frequencies seem to be adjoined rather closely in the visual representation, but, this shortcoming allows for better real-time resolution when performing sound analysis of EVP files.

“Seeing” Sounds

                In order to “read” the images of sounds that a spectrogram will create for us we must first understand the qualities of normal human speech so that we will have a baseline to compare our EVP recordings to. Before we get too far into this portion of the chapter I should mention that the sounds that are going to be analyzed will be those of a native speaker of North American English.

                Please note that sounds in all languages are organized into a sequence of abstract units called phonemes. For those of you who are not native speakers of North American English I would suggest researching a set of phonemes that you are accustomed to hearing before you try interpreting a language that is not your native tongue. Most times, web sites on speech and speech therapy can be found in your native language through a basic internet search. This research will make your understanding of the subject more complete and your EVP analysis more accurate.

                When analyzing speech patterns on a spectrogram the program will apply a mathematical technique called Fourier analysis to the acoustical sound waves so that we may know what frequency that particular waveform is on at any given time. Typically what occurs during this process is that the recording is played through the spectrogram and then broken down into very small pieces of data from ½ of a second to 1 second. These pieces of the whole are then analyzed to find what frequency they are on. Once the analysis of each of the data pieces is complete the results are added up and then divided by the number of data bits that were used. This gives us the mean frequency of the overall EVP.

                What you will find in the next few paragraphs is a basic lesson on the function of sounds within the speech of the North American English language. Only the primary categories of vowels and consonants will be covered so that we can, at a minimum, understand the contrast between the most basic speech sounds and patterns for analytical purposes. The seven categories of speech sounds are divided by their mode of phonation:

Consonants: Approximants, Nasals, Fricatives, Plosives and Affricates

                There are four (4) Approximants, also known as semi-vowels, which have sounds midway between a consonant and a vowel in the American English language. It has been found that in these phoneme categories there tends to be more constriction in the larynx for these vowels than for normal vowels, but conversely there is less tension for these consonants than in the other consonant categories.

                American English only has three (3) Nasals in which the air flow is completely blocked from the vocal tract. An example of this type of nasal can be found in the “ng” portion of the word “sing,” the “m” in the word me, and the “n” in the word new. The nasals have much less energy than many of the other phonation categories. This is because the oral tract is completely blocked, and sound waves radiate principally from the nose.

                The American English language has nine (9) Fricatives which are denoted by either weak or strong noises when produced, however this is the case only if the articulators are close enough together to cause a disruption of the air flow. These are “F”, “H”, “S”, “V”, “Z”, “th”, and “sh.” The “th” and “sh” account for two sounds a piece as they have both hard and soft pronunciations. The fricatives do not necessarily involve any voicing, although the voiced fricatives may have a very low voice bar. The signature of fricatives is in their high-frequency regions, which are more random in their energy distribution than voicing.

                There are six (6) Plosives in the American English language system and, as the name would imply, these are bursts or “explosions” of acoustic energy following a short period of silence; because of the silence during which the vocal tract is completely blocked, these phonemes are also called stops. The signature of plosives is an almost instantaneous passage from little or no acoustic energy to a short burst of high-energy in a wide frequency band. These sounds are created by a relatively complete closure of the vocal tract followed by a rapid release of the larynx to make “B”, “D”, “G”, “K”, “P”, and “T” sounds.    

                There are only two (2) Affricates in the American English language system which is really plosives that are released as if they were fricatives. These would be the sounds found in the “ch” portion of the word change and “dge” in the word grudge.

Vowels: Monophthongs & Diphthongs

                Monophthong vowels are characterized by strong stable voicing, and, speaking in terms of American English there are eleven (11) monophthong vowels which have a single vowel quality and two (2) that have a reduced quality. American English also has six (6) diphthongs which, unlike monophthongs have a strong but liquid voicing to them. These are vowels that generally manifest a clear change in quality from the start of the sound to the end of it. 

Common Sounds

                When interpreting a potential EVP with a spectrogram I will generally break down the various sounds into four categories: voicing, A/Nr noise, NVA noise and anomalous sounds. This is where the analysis of our potential EVP file will begin. We must first determine what sounds are found within the recording.

Category 1: Voicing

                As the name suggests, voicing sounds are just that – someone talking in either the foreground (usually the investigator holding the mic or recorder) or possibly someone in the background who is speaking. Either way this type of sound is generated by a first party individual who is in the area at the time the sound file is being recorded.

                This type of sound will show up on the spectrogram as strong vertical striations in a waveform analysis, as brightly hued bands in a color spectrogram, or as deep black areas on a black and white sonogram. Often times, when working with EVP that has voicing in it I will create a “voice bar,” that is a notation on the highest pitch and the lowest pitch of the voicing sounds as a gauge against other potentially anomalous sounds that I may find.

                In the section entitled “The Mechanics behind EVP” it was stated that the general accepted range of human vocal sounds is between 280 Hz and 1000 Hz, however the human ear can hear in ranges of 20 Hz to 20,000 Hz (20 kHz). Essentially, what this means is that humans have a far greater range and capability to hear than they do to produce sounds. Although sound production is “limited” there is still enough of a range within the human voice to perplex even some of the best EVP researchers at times. 

Category 2: A/Nr (Arrhythmic / Non-repetitive) Noise

                A/Nr noise is an arrhythmic short-term sound which is non-repetitive and can be attributed to a source that is generally mechanical in nature such as a passing car, a train horn, a plane in the distance, machinery, a door slamming, etc. Noises such as this are, for the most part, easily differentiated from any type of anomalous sound as they tend to have a mechanical sound to them rather than a voicing sound. A/Nr sounds, due to their brief nature are generally something that is unexpected and out of the control of the investigator. Although A/Nr sounds are generally not detrimental to the attempt to capture EVP they can be annoying when the EVP is being analyzed as they are one more thing to have to listen to and determine whether it is an anomaly or not.

Category 3: NVA (Non-Voicing / Ambient) Noise

                NVA is considered to be an aperiodic non-voicing or ambient background noise that can be caused by, but is in no way limited to: echo, wind, electrical hum, water, silence, etc. NVA is generally understood as “environmental” factors that investigators generally do not have control over. These, like A/Nr noises, are generally easy to distinguish from anomalous sounds since they tend to be in the background and are generally just noise rather than any type of organized, intelligible sounds.

Category 4: Anomalous Sounds

                These are sounds that do not fall into any of the previous three categories and can not generally be defined as such by logical, rational or analytical means by the investigator. Anomalous sounds, at times, can be very illusory due to the fact that they can sound either like voicing noises or A/Nr sounds, especially when dealing with Class A EVP. When analyzed however, it has been found that these sounds are not actually voicing or A/Nr sounds. When they are checked against biological factors and the limitations of the human voice it is generally found that the anomalous sounds are beyond the range of human capacity, and thusly will exceed the range of physical limitation, not to mention that the voice usually doesn’t sound like anyone who was on the investigation crew. Bear in mind though that just because we are generally listening for a vocal EVP that doesn’t always mean we will get one. In some cases the EVPs captured are of sounds, echoes in time of events that have happened in earlier years and, although these are not vocal EVPs, they are still EVPs nonetheless.

Understanding the Spectrographic Axis

Time – (X)

                Time, when speaking of spectrograms, generally doesn’t apply in the way that we understand it. Time on a spectrogram is called the “X axis,” and is generally found along the lower horizontal bar near the bottom of a spectrogram screen. Often this scale is broken down into both seconds and milliseconds. 

10.3 – This is a spectrogram with an active sound file. Note the timeline at the bottom as well as the “clock” below it.

Frequency – (Y)

                Frequency, also known as pitch, on a spectrogram is read from left to right and a typical spectrographic scale will run from zero (0) to about thirty thousand (30,000) cycles per second, with the “cycles per second” being interpreted in Hertz (Hz) or kilohertz (KHz) which is Hz x 1,000. The frequency or “Y” axis is usually found on a vertical bar to the right of the spectrogram screen. This is where the “energy” of the wave form is found.

                An accurate understanding of frequency in the analysis of EVP is of the utmost importance. To get a better understanding of frequency we can say that in any case where physical events are cyclical (or at least nearly cyclical) the frequency acts as a measurement for the intangible events that take place.

Frequency can be thought of like this:

T / C = F

T = Time, C = number of full cycles within given period of time, F = frequency.

                Basically what this means is that the number of cycles that complete in a given duration of time is the frequency of that event. So, if an event or sound occurs 120 times per minute it can be said that it has a frequency of 120 cycles per minute – or – 120 Hz = 1 longitudinal vibration per second.

10.4 – This photo of a spectrogram reveals an independent frequency analysis screen in the upper portion of the screen. Generally the frequency portion of the spectrogram is located on the right side vertical to the screen.

Amplitude (Decibels)

                Amplitude is the intensity of the sound which is measured in decibels and is depicted as the sharp vertical striations on a waveform analysis. On a spectrogram, however, amplitude is depicted by color saturation or by the color coding that the sounds produce. This information, in addition to the frequency that is produced by the sound will give the investigator a good idea of what kind of acoustical energy was involved in making the sounds heard on the potential EVP.

10.5 – This color spectrogram / sonogram shows the sound amplitude (intensity) through a varying degree of color.

A Final Note on Spectrograms

                And now for something truly obvious (which I cannot stress enough): When using sound enhancement/ analysis/ diagnostic software you should read the manual thoroughly and experiment with the program so that you have a feel for the tools you are using before you actually start modifying EVP files. The more familiar you are and the more you work with your spectrogram program the better you will get with it.