Audio monitoring system for assessing wildlife biodiversity

• US 5,956,463 A
• Filed: 10/07/1996
• Issued: 09/21/1999
• Est. Priority Date: 06/15/1993
• Status: Expired due to Fees

• Current Assignee: Kinectrics Incorporated
• Original Assignee: Ontario Hydro
• Inventors: Hanson, William G., Ramani, Narayan, Jennette, Robert L., Sheehan, Ronald W., Patrick, Paul H.
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): SAX, ROBERT L
• Application Number:
  US08/726,425
• Time in Patent Office: 1,079 Days
• Field of Search: 395/2.41, 395/2.55, 119/906
• US Class Current: 704/232
• CPC Class Codes: G10L 15/16 using artificial neural networks
• CPC Class Codes: G10L 17/26 Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices
• CPC Class Codes: G10L 2015/0631 Creating reference templates; Clustering
• CPC Class Codes: Y10S 119/906 Animal counting

Abstract

The invention relates to an automated system for monitoring wildlife auditory data and recording same for subsequent analysis and identification. The system comprises one or more microphones coupled to a recording apparatus for recording wildlife vocalizations in digital format. The resultant recorded data is preprocessed, segmented, and analyzed by means of a neural network to identify the respective species. The system minimizes the need for human intervention and subjective interpretation of the recorded sounds.

Claims

1. A method for assessing biodiversity in an environment, said method comprising the steps of:

obtaining vocalizations from a plurality of locations in the environment;

deriving auditory data from the vocalizations for each of said locations;

recording the auditory data in digital format;

processing the recorded data for each of said locations;

identifying predetermined characteristics of the processed data to identify the wildlife species from which the vocalizations are derived for each of said locations;

quantifying the wildlife species identified at each of said locations; and

assessing the biodiversity of the environment on the basis of the number of wildlife species identified.

2. The method for assessing biodiversity according to claim 1, further including the step of determining identified members of a wildlife species which move from one of said locations to another of said locations, so that the identified member is counted once in the quantification step.

3. An automated system for assessing biodiversity in an environment, said system comprising:

means for receiving vocalizations from a plurality of locations in the environment;

means for deriving auditory data from the vocalizations for each of said locations;

means for recording the auditory data in digital format;

means for processing the recorded data for each of said locations;

means for identifying predetermined characteristics of the processed data thereby to identify the wildlife species from which the vocalizations are derived for each of said locations;

means for quantifying the wildlife species identified at each of said locations and means for assessing the biodiversity of the environment on the basis of the identified wildlife species.

4. The automated system for assessing biodiversity according to claim 3, wherein said receiving means comprises at least one microphone located at each of said locations and coupled to the recording means.

5. The automated system for assessing biodiversity according to claim 4, wherein at least three locations are selected for the environment of interest and each of said locations includes at least one microphone, said microphone being coupled to said recording means.

6. The automated system for assessing biodiversity according to claim 4, wherein said recording means is time-triggered to record the auditory data at successive discrete intervals.

7. The automated system for assessing biodiversity according to claim 3, wherein said identifying means comprises:

(a) means for determining the commencement of a vocalization in said recorded auditory data;

(b) means for characterizing a received vocalization into a set of numerically quantized features of said vocalization using a characterizing technique; and

(c) means for classifying the vocalization according to said numerically quantized features.

8. The automated system for assessing biodiversity according to claim 7, wherein said means for assessing the biodiversity includes means for detecting members of an identified wildlife species moving from one of said locations to another of said locations.

9. An automated monitoring system for monitoring wildlife vocalizations comprising:

means for receiving and deriving auditory data from said vocalizations,

means for recording the auditory data in digital format,

means for processing the recorded data,

means for identifying predetermined characteristics of the processed data thereby to identify the wildlife species from which the vocalizations are derived, and

said means for identifying including means for determining the commencement of a vocalization in said recorded auditory data means for characterizing a received vocalization into a set of numerically quantized features of said vocalization using a characterizing technique, and means for classifying the vocalization according to said numerically quantized features.

10. An automated monitoring system according to claim 9, wherein said receiving means comprise at least one microphone connected directly to the recording means.

11. An automated monitoring system according to claim 9, wherein said receiving means comprise at least one microphone coupled to the recording means by a radiofrequency link.

12. An automated monitoring system according to claim 9, wherein said receiving means comprise at least three microphones located at different locations and coupled to the recording means by radiofrequency links defining respective receiving channels.

13. An automated monitoring system according to claim 9, wherein said recording means comprises an analog-to-digital converter connected to a personal computer.

14. An automated monitoring system according to claim 13, wherein the personal computer includes a hard disk.

15. An automated monitoring system according to claim 9, wherein said recording means is time-triggered to record the auditory data at successive discrete intervals of equal duration.

16. An automated monitoring system according to claim 9, wherein said recording means is sound-activated to record the auditory data at successive discrete intervals of equal duration.

17. An automated monitoring system according to claim 9, further comprising means for sensing environmental data, said recording means being adapted to record the environmental data in digital format.

18. An automated monitoring system according to claim 17, wherein said sensing means are temperature, barometric pressure and wind velocity sensors.

19. An automated monitoring system according to claim 9, wherein said processing means for processing the recorded data comprises means for formatting the data into digital data files and means for compressing the digital data files to a uniform size without information loss.

20. An automated monitoring system according to claim 19, wherein said processing means further comprises means for deriving from the recorded data spectrograms, audiograms and time domain representations.

21. An automated monitoring system according to claim 20, wherein said processing means further comprises means for deriving from the recorded data

(i) the average strength of all the frequency components at a particular time as a function of time,

(ii) the strength of the dominant frequency at a particular time as a function of time, and

(iii) the standard deviation for the portion of the auditory data containing noise and the portion of the auditory data representing the vocalization.

22. An automated monitoring system according to claim 20, wherein said processing means further comprises:

(a) means for detecting a vocalization using the standard deviation of said digitally recorded received sounds;

(b) means for filtering said audiogram so as to eliminate ambient noise or separate simultaneous vocalizations from species having vocalizations of different frequencies; and

(c) means for smoothing said audiogram and said time domain representation by averaging.

23. An automated monitoring system according to claim 9, further comprising a digital-to-analog converter connected to said processing means for converting said digitally recorded received sounds to analog form, and a speaker connected to receive said converted sounds thereby to enable auditory verification of the results of said identifying means.

24. An automated monitoring system according to claim 9, wherein said means for determining the commencement of said vocalization performs the following steps:

(a) dividing the digitally recorded data into a plurality of segments;

(b) calculating the acoustical energy of each segment;

(c) comparing the acoustical energy of each segment with a predetermined threshold value; and

(d) locating a predetermined number of contiguous segments whose acoustical energy exceed said threshold value.

25. An automated monitoring system according to claim 9, wherein said characterizing means characterizes said recorded auditory data using mel bins, cepstrum coefficients, linear predictive coefficients or correlation coefficients.

26. An automated monitoring system according to claim 9, wherein said classifying means comprises a neural network.

27. An automated monitoring system according to claim 26, wherein the neural network is a multilayer fully connected feed forward perception type having an input layer, an output layer and at least one hidden layer.

28. An automated monitoring system according to claim 27, wherein the number of output neurons in said input layer corresponds to the number of numerically quantized features in said set and the number of neurons in the output layer corresponds to the number of possible classifications for said vocalization.

29. An automated monitoring system according to claim 28, wherein the activation of each output neuron in said output layer identifies a further sub-classification neural network which has an input layer which receives said numerically quantized features and an output layer having a neuron for each possible classification for said vocalization.

30. An automated monitoring system according to claim 29, further comprising a successive classification neural network for use in further classifying said vocalization.