Error protection for multimode speech coders

• US 5,642,368 A
• Filed: 10/11/1995
• Issued: 06/24/1997
• Est. Priority Date: 09/05/1991
• Status: Expired due to Term

• Current Assignee: Google Technology Holdings LLC (Alphabet Inc.)
• Original Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Inventors: Winter, Eric Helmut, Gerson, Ira Alan, Jasiuk, Mark Antoni
• Primary Examiner(s): Snyder, Glenn
• Application Number:
  US08/541,738
• Time in Patent Office: 622 Days
• Field of Search: 371/37.1, 371/43, 371/2.1, 381/36, 381/46, 395/2.09, 395/2.17, 395/2.23, 395/2.28, 395/2.42
• US Class Current: 714/786
• CPC Class Codes: G10L 19/06 Determination or coding of the spectral characteristics, e.g. of the short-term prediction coefficients
• CPC Class Codes: G10L 19/135 Vector sum excited linear prediction [VSELP]
• CPC Class Codes: G10L 19/18 Vocoders using multiple modes
• CPC Class Codes: H03M 13/35 Unequal or adaptive error protection, e.g. by providing a different level of protection according to significance of source information or by adapting the coding according to the change of transmission channel characteristics

Abstract

A method for protecting information bits wherein input data bits, at least some of which are to be protected, are sorted based upon information determined from a subset of the input data bits. An error control coding technique is applied to at least some of the sorted bits. In the preferred embodiment, an input data stream of voice coder bits is separated into arrays of bits. A first array (302) comprises voice coder bits needing error protection, with the bits arranged in order of importance determined by voicing mode. The second array (303) comprises bits that will not be error protected. The bits from the first array are provided to the input of an encoder (304), then the encoded bits are combined (305) with the bits from the second array (303) to form a bit stream.

Claims

1. A speech coder for producing coded speech, the speech being one of voiced speech and unvoiced speech, the speech coder comprising:

a linear predictive coding (LPC) filter having an input for receiving an excitation and an output for providing a speech approximation signal, the speech approximation signal including data bits representative of voiced speech;

a comparator having an input coupled to the LPC filter output for receiving the speech approximation signal and an output, the comparator comparing the speech approximation signal with digitized speech and providing an error signal representative of the difference between the digitized speech and the speech approximation signal to the output;

a long-term predictor having an input coupled to the comparator and an output, the long term predictor providing a long term predictor vector at the output in response to the error signal received at the input only when the speech is voiced speech;

one or more codebooks, each having a respective input and a respective output for providing a respective codebook vector, the one or more codebooks providing one respective codebook vector from one codebook of the one or more codebooks in response to the error signal when the speech is voiced speech, the one or more codebooks providing more than one codebook vector from more than one respective codebooks in response to the error signal when the speech is unvoiced speech;

a summer having a plurality of inputs coupled to the long-term predictor output and the respective outputs of the codebooks, the summer summing the long-term predictor vector and the respective codebook vectors to produce the excitation, the summer having an output coupled to the input of the LPC filter for providing the excitation to the LPC filter;

an encoder circuit coupled to the output of the LPC filter for receiving the speech approximation signal, the encoder circuit including;

a data separator having an input for receiving the speech approximation signal and an output, the data separator separating the data bits representative of voiced speech into more important bits and less important bits in response to the speech being voiced speech;

a first memory having an input coupled to the output of the data separator for receiving the more important bits and an output, the first memory storing the more important bits;

a second memory having an input coupled to the output of the data separator for receiving the less important bits and an output, the second memory storing the less important bits;

a convolutional encoder having an input coupled to the output of the first memory and an output, the convolutional encoder encoding the more important bits to produce encoded data bits at the output; and

an interleaver having a first input coupled to the output of the convolutional encoder for receiving the encoded data bits, a second input coupled to the output of the second memory for receiving the less important bits and an output, the interleaver combining the encoded data bits and the less important data bits to produce the coded speech.

2. A speech coding system as defined in claim 1 wherein the encoder comprises a multirate convolutional encoder.

3. A speech coding system as defined in claim 1 wherein the encoder comprises a convolutional encoder variable from a rate 1/3 encoder to a rate 1/2 encoder.

4. A speech coding system as defined in claim 1 wherein the most important bits comprise bits defining the voicing mode of the speech.

5. A speech coding system as defined in claim 4 wherein the encoder encodes the most important data bits serially, starting with a first bit to be encoded and ending with a last bit to be encoded, and wherein the bits defining the voicing mode are encoded near the first bit to be encoded.

6. A method for producing coded speech, the speech being one of voiced speech and unvoiced speech, the method comprising the steps of:

producing a speech approximation signal, the speech approximation signal including data bits representative of voiced speech;

comparing the speech approximation signal with digitized speech and providing an error signal representative of the difference between the digitized speech and the speech approximation signal;

providing a long term predictor vector in response to the error signal only when the speech is voiced speech;

selecting a codebook vector from a codebook in response to the error signal when the speech is voiced speech, and selecting more than one codebook vector in response to the error signal when the speech is unvoiced speech;

separating the data bits representative of voiced speech into more important bits and less important bits when the speech is voiced speech;

storing the more important bits in a first memory;

storing the less important bits in a second memory;

encoding the more important bits to produce encoded data bits; and

combining the encoded data bits and the less important data bits to produce the coded speech.

7. A method as recited in claim 6 further comprising the step of summing the long term predictor vector and the codebook vector, producing an excitation, wherein the speech approximation signal is produced responsive to the excitation.

8. A method as recited in claim 6 wherein the step of encoding comprises the steps of encoding the most important data bits serially, starting with a first bit to be encoded and ending with a last bit to be encoded, and encoding bits defining the voicing mode near the first bit to be encoded.