New Measurements

Introduction

A number of new measurements were added when we released Raven Pro 1.5 which do not appear in the Raven Pro User’s Manual.

Frequency Contour Measurements

Definition: Measurements calculated by applying a Raven frequency measurement to each spectrogram slice

• Frequency Contour Percentile 5%
  Series of numbers consisting of a Frequency 5% measurement for each spectrogram slice
• Frequency Contour Percentile 25%
  Series of numbers consisting of a 1st Quartile Frequency measurement for each spectrogram slice
• Frequency Contour Percentile 50%
  Series of numbers consisting of a Center Frequency measurement for each spectrogram slice
• Frequency Contour Percentile 75%
  Series of numbers consisting of a 3rd Quartile Frequency measurement for each spectrogram slice
• Frequency Contour Percentile 95%
  Series of numbers consisting of a Frequency 95% measurement for each spectrogram slice

More

Peak Frequency Contour Measurements

• Peak Frequency Contour
  Series of numbers consisting of a Peak Frequency measurement for each spectrogram slice
• Peak Frequency Contour Slope
  Series of numbers of consisting of the difference between each pair of adjacent elements in Peak Frequency Contour
• Peak Frequency Contour Average Slope
  Mean of the Peak Frequency Contour Slope Series of numbers
• Peak Frequency Contour Max Frequency
  Maximum of the Peak Frequency Contour Series of numbers
• Peak Frequency Contour Max Slope
  Maximum of the Peak Frequency Contour Slope Series of numbers
• Peak Frequency Contour Min Frequency
  Minimum of the Peak Frequency Contour Series of numbers
• Peak Frequency Contour Min Slope
  Minimum of the Peak Frequency Contour Slope Series of numbers
• Peak Frequency Contour Number of Inflection Points
  Number of times the slope changes sign in Peak Frequency Contour Slope Series of numbers

More

Entropy Measurements

• Minimum Entropy
  The minimum entropy calculated for a spectrogram slice within the selection bounds.

• Maximum Entropy
  The maximum entropy calculated for a spectrogram slice within the selection bounds.

Entropy in Raven is computed on a per-spectrogram-slice basis. The formula is

S = PSD(f,t)/sum_over_f(PSD(f,t)) * log2( PSD(f,t)/sum_over_f(PSD(f,t)))

The units are “bits” because we use the log base 2. Since the selection may consist of multiple spectrogram slices, Raven iterates over slices and to find the minimum and maximum entropy value with the frequency bounds of the selection. Note that most signal processing applications sum over frequency and time, where Raven sums over frequency instead.

Relative Time Measurements

• Time 5% Relative
  Proportion of the selection at which 5% of the sound energy has an earlier time

• Time 25% Relative
  Proportion of selection at which 25% of the sound energy has an earlier time

• Center Time Relative
  Proportion of selection at which 50% of the sound energy has an earlier time

• Time 75% Relative
  Proportion of selection at which 75% of the sound energy has an earlier time

• Time 95% Relative
  Proportion of selection at which 95% of the sound energy has an earlier time

• Peak Time Relative
  In a waveform view, the first time in a selection at which a sample with amplitude equal to Peak Amplitude occurs, the time expressed as a proportion of the duration of the selection. In a spectrogram view, the first time in a selection at which a spectrogram bin with power density equal to Peak Power occurs, the time expressed as a proportion of the duration of the selection.

Power Measurements

Unless you have calibrated a sound view Raven’s power, energy, and amplitude measurements are relative to an arbitrary reference, meaning that you can make relative comparisons between measurements in sound recordings made with the same recording equipment and settings, but *not* absolute measurements. For more about Raven calibration, see Calibration.

• Inband Power
  Integral of the average power spectral density (PSD) over the frequency bounds of each selection. Since Raven uses discrete Fourier transform (DFT), it averages the power spectral density with respect to time, sums up the time-averaged PSD in all DFT bins and multiplies by the bin size in Hertz.
• Leq
  Equivalent continuous sound level (Leq) is the preferred method to describe sound levels that vary over time, resulting in a single decibel value which takes into account the total sound energy over the period of time of interest. Raven uses slow time weighting and no frequency weighting.
• SEL
  Sound Exposure Level (dB) is the log10 integral, over time, of sound pressure relative to a reference pressure. Note that frequency weighting is not employed by Raven.

Power Density Measurements

Several changes were made in Raven Pro 1.6 that impact how Raven reports power density measurements.

• The word “density” was added to measurement names to make it clearer that they are power density measurements.
  • Average Power became Average Power Density.
  • Delta Power became Delta Power Density.
  • Maximum Power and Peak Power became Peak Power Density.
• Power density measurements are now reported in units of dB/Hz. This is a more customary unit for power density since it facilitates the comparison of values across spectrogram parameters. In Raven Pro 1.6 power spectral density (dB/Hz) is equal to the Raven Pro 1.5 value (dB) plus 10 times the base 10 logarithm of the DFT size used to generate the spectrogram.
• In Raven Pro 1.6 the default unit for reporting power and power density measurements is dB FS, in which zero is defined as the highest sound amplitude that can be recorded with your recording system in its current configuration. Our hope is that this convention will make it clearer that by default Raven reports uncalibrated power and power density measurements. Once the user has calibrated a signal in Raven Pro 1.6, power and power density are reported in decibels relative to a known reference pressure (typically 20 micropascals in air and 1 micropascal in water). Information about calibrating signals in Raven can be found in Power and Amplitude Measurements.

Power density measurements are not the same as power measurements such as Inband Power. For example, Average Power Density is the sum of the square magnitudes of the Fourier coefficients across time and frequency, divided by the the product of the number of Fourier bins in the selection and the number of spectrogram frames in the selection. Another way to think of it is the sum of the square magnitude across time and frequency, divided by the product of the selection duration (Delta Time) and the selection bandwidth (Delta Frequency). Conceptually, this really is the 2D average of the square magnitudes.

 The following exercise may help you get an intuitive sense of the difference between a power measurement and a power spectral density measurement.

1. Find a part of your recording where most of the sound energy is concentrated in a narrow frequency band, and make a selection around it. The sound file named 2000Hz.aif in the Examples folder works well for this demonstration.
2. Add the Inband Power and Average Power Density measurements to your selection table.
3. Slowly increase the frequency range (that is, bandwidth) of the selection you just made, and observe how the values of Inband Power and Average Power Density change.
   • Inband Power, if it changes, will only get bigger because your selection encloses more power as you make your selection taller.
   • Average Power Density, if it changes, will generally decrease because you are adding spectrogram bins with low power spectral density; In other words, if you are using the default “Grayscale” colormap, the average darkness of the spectrogram you are adding as you increase the height of your selection decreases.

Selection Boundary Measurements

• Sample Length Selection
  Selection duration in samples (i.e. Delta Time x Sample Rate)

Real Clock Times Measurements

Real Clock Times set up is discussed in Real Clock Times.

• Begin Clock Time
  Time the beginning of the selection was originally recorded (eg, 18:22:59.7973)

• Begin Date
  Date the beginning of the selection was originally recorded (eg, 2020/4/9)

• Begin Date Time
  Date and time the beginning of the selection was originally recorded (eg, 2020/4/9  18:22:59.7973)

• Begin Hour
  Hour the beginning of the selection was originally recorded (eg, 18)

• End Clock Time
  Time the end of the selection was originally recorded (eg, 18:27:44.1531)

• End Date
  Date the end of the selection was originally recorded (eg, 2020/4/9)

Signal To Noise Ratio

• Signal-to-Noise Ratio (SNR) – NIST Quick method
• Signal-to-Noise Ratio (SNR) – user protocol

Glossary

• Selection bounds
  The time and frequency limits defining a selection; in other words, Begin Time, End Time, Low Frequency and High Frequency

• Energy
  The capacity to do work, in the case of sound the capacity to vibrate particles in the air, water, or some other medium

• Power
  Energy per unit time

• PSD
  Energy per unit frequency per unit time. Each color in a Raven spectrogram represents a range of PSD.

• Spectrogram bin
  One value in a spectrogram matrix. Turning off spectrogram smoothing in Raven allows Raven to display spectrogram bins as colored rectangles, which are more visible when you zoom in.

• Spectrogram slice
  A column of spectrogram bins the width of the Hop Size and spanning the frequency range from 0 Hz to the Nyquist frequency.

• Frequency slice
  A row of spectrogram bins the height of the frequency grid interval spanning the spectrogram.

• Percentile
  The value below which a given percentage of bins  falls

• Quartile
  The 25th percentile, median, and 75th percentile