Compressive sensing for full matrix capture
Examples of the present subject matter provide techniques for compressive sampling of acoustic data. A probe may sample in a compression mode, such that the entire matrix is not sampled at full-time resolution or spatial resolution. Therefore, the initial amount of data captured by the probe is reduced, allowing for lower density hardware (e.g., fewer analog-to-digital conversion channels or related analog front-end hardware) to be used at a lower data rate.
1. A method comprising:
obtaining signals representative of one or more acoustic waves received using an acoustic probe with a matrix of sensing elements in response to one or more acoustic signals being transmitted into an object under test using the acoustic probe, wherein N samples define a full matrix representation of the signals;
under sampling the obtained signals by a subset M samples, wherein N is greater than M;
transmitting the M samples to a post-acquisition application; and
reconstructing, by the post-acquisition application, an image from the M samples, wherein reconstructing the image comprises:
establishing a model of a full-matrix representation;
based on the model, converting the M samples to a reconstructed full-matrix representation of the signals; and
generating a reconstructed image based on the reconstructed full-matrix representation, wherein the model includes an inverse sparse matrix, and coefficients of the inverse sparse matrix are based on characteristics of the object under test.
2. The method of claim 1 , further comprising:
applying weights to the M samples.
3. The method of claim 1 , wherein the M samples are randomly selected.
4. The method of claim 1 , wherein coefficients of the inverse sparse matrix are based on characteristics of the matrix of sensing elements.
5. A non-transitory machine-storage medium embodying instructions that, when executed by a machine, cause the machine to perform operations comprising:
obtaining signals representative of one or more acoustic waves received using an acoustic probe with a matrix of sensing elements in response to one or more acoustic signals being transmitted into an object under test using the acoustic probe, wherein N samples define a full matrix representation of the signals;
under sampling the obtained signals by a subset M samples, wherein N is greater than M;
transmitting the M samples to a post-acquisition application; and
reconstructing, by the post-acquisition application, an image from the M samples, wherein reconstructing the image comprises:
establishing a model of a full-matrix representation;
based on the model, converting the M samples to a reconstructed full-matrix representation of the signals; and
generating a reconstructed image based on the reconstructed full-matrix representation, wherein the model includes an inverse sparse matrix, and coefficients of the inverse sparse matrix are based on characteristics of the object under test.
6. The non-transitory machine-storage medium of claim 5 , further comprising:
applying weights to the M samples.
7. The non-transitory machine-storage medium of claim 5 , wherein the M samples are randomly selected.
8. The non-transitory machine-storage medium of claim 5 , wherein coefficients of the inverse sparse matrix are based on characteristics of the matrix of sensing elements.
9. A system comprising:
one or more processors of a machine; and
a memory storing instructions that, when executed by the one or more processors, cause the machine to perform operations comprising:
obtaining signals representative of one or more acoustic waves received using an acoustic probe with a matrix of sensing elements in response to one or more acoustic signals being transmitted into an object under test using the acoustic probe, wherein N samples define a full matrix representation of the signals;
under sampling the obtained signals by a subset M samples, wherein N is greater than M;
transmitting the M samples to a post-acquisition application; and
reconstructing, by the post-acquisition application, an image from the M samples, wherein reconstructing the image comprises:
establishing a model of a full-matrix representation;
based on the model, converting the M samples to a reconstructed full-matrix representation of the signals; and
generating a reconstructed image based on the reconstructed full-matrix representation, wherein the model includes an inverse sparse matrix, and coefficients of the inverse sparse matrix are based on characteristics of the object under test.
10. The system of claim 9 , the operations further comprising:
applying weights to the M samples.
11. The system of claim 9 , wherein the M samples are randomly selected.
12. The system of claim 9 , wherein coefficients of the inverse sparse matrix are based on characteristics of the matrix of sensing elements.