Real-time traffic detection

• US 9,424,743 B2
• Filed: 10/10/2013
• Issued: 08/23/2016
• Est. Priority Date: 10/12/2012
• Status: Active Grant

• Current Assignee: TATA Consultancy Services Limited (Tata Sons Pvt Ltd.)
• Original Assignee: TATA Consultancy Services Limited (Tata Sons Pvt Ltd.)
• Inventors: Banerjee, Rohan, Sinha, Aniruddha
• Primary Examiner(s): ODOM, CURTIS B
• Application Number:
  US14/431,053
• Publication Number:
  US 20150248834A1
• Time in Patent Office: 1,048 Days
• US Class Current: 1/1
• CPC Class Codes: G08G 1/01 Detecting movement of traffic to be counted or controlled G08G1/07 - G08G1/14 take precedence
• CPC Class Codes: G08G 1/0104 Measuring and analyzing of parameters relative to traffic conditions
• CPC Class Codes: G08G 1/0133 for classifying traffic situation
• CPC Class Codes: G08G 1/04 using optical or ultrasonic detectors

Abstract

Systems and methods for real-time traffic detection are described. In one embodiment, the method comprises capturing ambient sounds as an audio sample in a user device, and segmenting the audio sample into a plurality of audio frames. Further, the method comprises identifying periodic frames amongst the plurality of audio frames. Spectral features of the identified periodic frames are extracted, and horn sounds are identified based on the spectral features. The identified horn sounds are then used for real-time traffic detection.

Claims

1. A method for real-time traffic detection, wherein the method comprising:

capturing ambient sounds as an audio sample in a user device;

segmenting the audio sample into a plurality of audio frames;

identifying periodic frames amongst the plurality of audio frames, wherein the identifying comprises separating the plurality of audio frames into the periodic frames, non-periodic frames, and silenced frames based on a short term energy level (En) and a Power Spectral Density (PSD) of the plurality of audio frames; and

extracting spectral features of the periodic frames for real-time traffic detection.

2. The method as claimed in claim 1, wherein the ambient sounds include one or more of tire noise, horn sound, engine noise, human speech, and background noise.

3. The method as claimed in claim 1, wherein the separating comprises

computing the short term energy level (En) for the plurality of audio frames;

and

comparing the short term energy level (En) of each of the plurality of audio frames with a predefined energy threshold to identify the silenced frames amongst the plurality of audio frames;

calculating a ratio of a maximum power spectral density and a total power spectral density (PSD) of remaining audio frames, wherein the remaining audio frames exclude the silenced frames; and

identifying the periodic frames amongst the remaining audio frames based on comparing the ratio of the maximum power spectral density and the total power spectral density with a predefined density threshold.

4. The method as claimed in claim 1 further comprising filtering background noise from the plurality of audio frames.

5. The method as claimed in claim 1, wherein the spectral features include one or more of Mel-Frequency Cepstral Coefficients (MFCC), inverse MFCC, and modified MFCC.

6. A method for real-time traffic detection, wherein the method comprising:

receiving spectral features of periodic frames from a plurality of user devices in a geographical location, wherein the periodic frames are identified based on a short term energy level (En) and a Power Spectral Density (PSID) of the plurality of audio frames;

identifying horn sounds based on the spectral features; and

detecting real-time traffic congestion at the geographical location based on the horn sounds.

7. The method as claimed in claim 6, wherein the spectral features include one or more of Mel-Frequency Cepstral Coefficients (MFCC), inverse MFCC, and modified MFCC.

8. The method as claimed in claim 6, wherein the identifying is based on at least one sound model, wherein the at least one sound model is any one of a horn sound model and a traffic sound model.

9. A user device for real-time traffic detection comprising:

a device processor; and

a device memory coupled to the device processor, the device memory comprising;

a segmentation module configured to segment an audio sample captured in the user device into a plurality of audio frames;

a frame separation module configured to separate the plurality of audio frames into at least periodic frames and non-periodic frames, wherein the frame separation module is configured to separate the plurality of audio frames based on a short term energy level (En) and a Power Spectral Density (PSD) of the plurality of audio frames; and

an extraction module configured to extract spectral features of the periodic frames, wherein the spectral features are transmitted to a server for real-time traffic detection.

10. The user device as claimed in claim 9, wherein the user device further comprising a filtration module configured to filter background noise from the plurality of audio frames.

11. A server for real-time traffic detection comprising:

a server processor; and

a server memory coupled to the server processor, the server memory comprising;

a sound detection module configured to;

receive spectral features of periodic frames from a plurality of user devices in a geographical location, wherein the periodic frames are identified based on a short term energy level (En) and a Power Spectral Density (PSD) of the plurality of audio frames; and

identify horn sounds based on the spectral features; and

a traffic detection module configured to detect real-time traffic congestion at the geographical location based on the horn sounds.

12. The server as claimed in claim 11, wherein the sound detection module is configured to identify the horn sounds based on at least one of a horn sound model and a traffic sound model.

13. A non-transitory computer-readable medium having embodied thereon a computer program for executing a method comprising:

capturing ambient sounds as an audio sample;

segmenting the audio sample into a plurality of audio frames;

identifying periodic frames amongst the plurality of audio frames, wherein the identifying comprises separating the plurality of audio frames into the periodic frames, non-periodic frames, and silenced frames based on a short term energy level (En) and a Power Spectral Density (PSD) of the plurality of audio frames;

extracting spectral features of the periodic frames;

identifying horn sounds based on the spectral features; and

detecting real-time traffic congestion based on the horn sounds.