Phoneme score accelerator

• US 9,514,739 B2
• Filed: 12/21/2012
• Issued: 12/06/2016
• Est. Priority Date: 06/06/2012
• Status: Active Grant

• Current Assignee: Cypress Semiconductor Corporation (Infineon Technologies AG)
• Original Assignee: Cypress Semiconductor Corporation (Infineon Technologies AG)
• Inventors: Fastow, Richard M., Bapat, Ojas A., Olson, Jens
• Primary Examiner(s): Sirjani, Fariba
• Application Number:
  US13/725,260
• Publication Number:
  US 20140180694A1
• Time in Patent Office: 1,446 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 15/28 Constructional details of speech recognition systems
• CPC Class Codes: G10L 2015/025 Phonemes, fenemes or fenones being the recognition units

Abstract

Embodiments of the present invention include an acoustic processing device and a method for traversing a Hidden Markov Model (HMM). The acoustic processing device can include a senone scoring unit (SSU), a memory device, a HMM module, and an interface module. The SSU is configured to receive feature vectors from an external computing device and to calculate senones. The memory device is configured to store the senone scores and HMM information, where the HMM information includes HMM IDs and HMM state scores. The HMM module is configured to traverse the HMM based on the senone scores and the HMM information. Further, the interface module is configured to transfer one or more HMM scoring requests from the external computing device to the HMM module and to transfer the HMM state scores to the external computing device.

Claims

1. An acoustic processing device for traversing a Hidden Markov Model (HMM), the device comprising:

a senone scoring unit (SSU) configured to receive feature vectors over a physical data bus from a computing device external to the acoustic processing device and to calculate senone scores;

a memory device configured to store the senone scores and HMM information, wherein the HMM information comprises HMM IDs and HMM state scores;

an HMM module configured to traverse the HMM based on the senone scores and the HMM information while the computing device searches for additional HMMs, wherein the HMM module comprises a buffer, an HMM generator, a histogram pruning module, an HMM scoring module, a histogram generator, a pre-pruning module, and an arc generator, wherein the buffer, the HMM generator, the histogram pruning module, the HMM scoring module, the histogram generator, the pre-pruning module, and the arc generator traverse the HMM in a pipelined manner, wherein pruning by the histogram pruning module is performed after the HMM generator, and wherein pre-pruning by the pre-pruning module is performed after the pruning by the histogram pruning module and before analysis of a subsequent frame of data; and

an interface module configured to transfer one or more HMM scoring requests associated with the HMM and additional HMMs from the computing device to the HMM module over the physical data bus and to transfer the HMM state scores to the computing device over the physical data bus.

2. The acoustic processing device of claim 1, wherein at least one of the HMM and the additional HMMs has a variable length.

3. The acoustic processing device of claim 1, wherein the SSU, the HMM module, and the interface module are integrated on a first chip and the memory device is integrated on a second chip.

4. The acoustic processing device of claim 1, wherein the HMM module is configured to evaluate one or more active lists that include the HMM information.

5. The acoustic processing device of claim 1, wherein the SSU is configured to calculate the senone scores based on an acoustic model.

6. The acoustic processing device of claim 1, wherein the SSU is configured to calculate the senone scores for a frame of data while the computing device searches for the additional HHMs.

7. The acoustic processing device of claim 1, wherein the HMM module is configured to evaluate new HMM information associated with a patch list, wherein the new HMM information is different from the HMM information stored in the memory device.

8. The acoustic processing device of claim 1, wherein the SSU is configured to calculate the senone scores prior to the calculation of the HMM state scores by the HMM module.

9. The acoustic processing device of claim 1, wherein the HMM module comprises:

the buffer configured to store new HMM information, wherein the new HMM information comprises new HMM IDs and new HMM state scores, and to transfer the HMM state scores to the computing device over the physical data bus;

the HMM generator configured to receive the new HMM information and reformat the new HMM information as one or more HMM states;

the histogram pruning module configured to apply a pruning algorithm to one or more HMM states to generate one or more active HMM states;

the HMM scoring module configured to generate the HMM state scores corresponding to the one or more active HMM states;

the histogram generator configured to generate a distribution of HMM state scores for calculation of a pruning threshold;

the pre-pruning module configured to designate the one or more active HMM states as inactive if the corresponding one or more HMM state scores is below a pre-pruning threshold; and

the arc generator configured to determine whether each of the one or more active HMM states with HMM state scores above the pre-pruning threshold is the last state of an HMM.

10. The acoustic processing device of claim 1, wherein the interface module is configured to transfer the one or more feature vectors from the computing device to the SSU unit.

11. A method for traversing a Hidden Markov Model (HMM), the method comprising:

receiving feature vectors from a computing device external to an acoustic processing device over a physical data bus;

calculating, with a senone scoring unit (SSU), senone scores;

storing, in a memory device, the senone scores and HMM information, wherein the HMM information comprises HMM IDs and HMM state scores;

applying, with an HMM module, the senone scores and the HMM information when traversing the HMM while the computing device searches for additional HMMs, wherein the applying comprises buffering, HMM generation, histogram pruning, HMM scoring, histogram generation, pre-pruning, and arc generation that are performed in a pipeline manner to traverse the HMM, wherein the histogram pruning includes applying a pruning algorithm after the HMM generation, and wherein the pre-pruning occurs after the histogram pruning and includes designating one or more HMM state as inactive before analysis of a subsequent frame of data; and

transferring one or more HMM scoring requests associated with the HMM and the additional HMMs from the computing device to the HMM module over the physical data bus and the HMM state scores to the computing device over the physical data bus.

12. The method of claim 11, wherein the applying of the HMM module comprises evaluating one or more active lists that include the HMM information.

13. The method of claim 11, wherein the calculating of the senone scores comprises calculating the senone scores based on an acoustic model.

14. The method of claim 11, wherein the calculating of the senone scores comprises calculating the senone scores for a frame of data while the computing device searches for the additional HMMs.

15. The method of claim 11, wherein the applying of the HMM module comprises evaluating new HMM information associated with a patch list, wherein the new HMM information is different from the HMM information stored in the memory device.

16. The method of claim 11, wherein the calculating of the senone scores comprises calculating the senone scores prior to the calculation of the HMM state scores by the HMM module.

17. The method of claim 11,

wherein the buffering comprises storing of new HMM information, wherein the new HMM information comprises new HMM IDs and new HMM state scores;

wherein the HMM generation comprises receiving and reformatting of the new HMM information as one or more HMM states;

wherein the histogram pruning comprises applying a pruning algorithm to one or more HMM states to generate one or more active HMM states;

wherein the HMM scoring comprises generating the HMM state scores corresponding to the one or more active HMM states;

wherein the histogram generation comprises generating a distribution of HMM state scores for calculation of a pruning threshold;

wherein the pre-pruning comprises designating the one or more active HMM states as inactive if the corresponding one or more HMM state scores is below a pre-pruning threshold; and

wherein the arc generation comprises determining whether each of the one or more active HMM states with HMM state scores above the pre-pruning threshold is the last state of an HMM.