JPH06104849A - Voice sampler - Google Patents

Abstract

PURPOSE:To provide the voice sampler using a dynamic range of a voice coding and decoding device fully and implementing voice sampling without distortion. CONSTITUTION:A voice input section of the voice sampler is provided with an A/D converter 12 having a higher resolution than that of input output data of a voice coder 16 and a voice decoder 18, after voice data are sampled and stored once in a memory 13, the resolution of the data is converted into a resolution in matching with the voice coder 16 in the input stage to the voice coder 16. Thus, the dynamic range of the resolution of input output data in the voice coding device and a voice decoding device is fully used and no distortion is caused to sample the voice signal.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech sampler, and more particularly to a speech sampler which makes full use of the dynamic range of a speech coder / decoder and is capable of distortion-free speech sampling.

[0002]

2. Description of the Related Art Conventionally, there is an audio sampler of this type as shown in FIG. This audio sampler includes an audio input terminal 1 for inputting an audio signal, an M-bit A / D converter 2 for converting an analog signal of an input audio into digital data with a resolution of M bits, and an M-bit length. Linear PCM audio data is input to encode the M-bit length word input audio encoder 3, a code memory 4 that stores the encoded audio data, and encoded audio data is an M bit long linear encoder. M-bit length word output voice decoder 5 for decoding into PCM voice data, M-bit D / A converter 6 for converting digital data with M-bit resolution into an analog signal, and voice for outputting voice signal An output terminal 7, an operation unit 8 having keys for inputting various commands, and a control unit 9 for performing integrated control of these are provided, and voice sampling is started via the operation unit 8. Then, the analog audio signal input through the audio input terminal 1 is A / D converted by the M-bit A / D converter 2 with M-bit resolution and converted into linear PCM audio data of M-bit length. After that, this voice data is input to the M-bit long word input voice encoder 3 and encoded, and then stored in the code memory 4 for voice sampling, and the voice is reproduced via the operation unit 8. Upon activation, the code stored in the code memory 4 is read out and input to the M-bit length word output voice decoder 5 to decode the voice, and the restored linear P of M-bit length is output.
The CM voice data is input to the M-bit D / A converter 6 to reproduce the sampled voice.

Here, M is a positive integer, and the encoding / decoding algorithm in the speech encoder 3 / speech decoder 5 is, for example, the ADPCM system, but the linear P
The present invention is not limited to this as long as the CM voice data is input and output, and the PCM method of storing the linear PCM voice data itself in the code memory is also included. Where P
CM is Pulse Code Modulatio
n is shown. In addition, ADPCM is Adaptive D
differential Pulse CodeMod
indicates the ration.

[0004]

However, in such a conventional voice sampling device, the resolution of the A / D converter 2 at the time of voice input is set in the voice encoder 3 and the voice decoder 5. Since the resolution is the same as the resolution of the output data, the S / N ratio decreases when the amplitude of the input audio signal is small, and distortion occurs when the amplitude of the input audio signal is large. 3 and speech decoder 5
There is a problem that it is difficult to sample the audio signal so that the dynamic range of the resolution of the input / output data is fully used and the distortion does not occur.

Further, in the conventional voice sampling apparatus, when the difference between the amplitude in the + direction and the amplitude in the-direction of the input voice is large, one of the large + or-amplitude is used in order to avoid distortion due to excessive input. Since the input level must be set according to the input level in the direction, the dynamic range of the resolution of the input / output data in the voice encoder 3 and the voice decoder 5 can be used only in one direction of + or −. , The voice encoder 3 and the voice decoder 5 can fully utilize the dynamic range of the resolution of input / output data in both the + direction and the-direction, and can sample a voice signal so that no distortion occurs. However, there is a problem that the S / N ratio is lowered or distortion is generated.

The present invention has been made in view of the above problems, and an object thereof is to provide an audio sample apparatus which can fully use the dynamic range of an audio encoder / decoder and can perform audio sampling without distortion. It is to be.

[0007]

In order to achieve the above-mentioned object, the present invention provides a voice sampling device for A / D converting an analog voice signal with N-bit resolution, and the A / D converter.
A storage means for storing N-bit resolution data output from the converter, and a resolution conversion means for converting data from N-bit resolution to M-bit resolution by multiplying N-bit resolution data by a constant. The main point is the provision.

According to a second aspect of the present invention, means for A / D converting an analog voice signal with N-bit resolution, and N-bit resolution data output from the A / D converter are provided to a voice sampling device. The storage means for storing the data and the N-bit resolution data stored in the storage means are converted so that the maximum amplitude becomes substantially the same in the ± direction of the amplitude, and then the data conversion from the N-bit resolution to the M-bit resolution is performed. The gist is that a voice data conversion means is provided.

[0009]

With the above construction, the input audio signal is first A / D converted by the A / D conversion means with N-bit resolution and temporarily stored in the storage means. After that, the A / D-converted N-bit resolution data is multiplied by a certain constant, and the data is converted from the N-bit resolution to the M-bit resolution. Then, the data after this resolution conversion is encoded by the encoding means. In the second invention of the present application, the input audio signal is first A / D converted by the A / D conversion means with N-bit resolution and temporarily stored in the storage means. Thereafter, this data is converted so that the maximum amplitude becomes substantially the same in the ± direction of the amplitude of the means, and the data conversion from N-bit resolution to M-bit resolution is performed. Then, the data after this resolution conversion is encoded by the encoding means.

[0010]

1 and 2 are views showing an embodiment of a voice sampling apparatus according to the first invention of the present application. In FIG. 1, reference numeral 11 is an audio input terminal for inputting an audio signal, 12 is an N-bit A / D converter which is a means for converting an analog signal of an input audio into digital data with N-bit resolution, and 13 is An A / D data storage memory for storing A / D-converted audio data of N-bit resolution, and 14 detects data having the maximum absolute value among the data stored in the A / D data storage memory 13. Maximum absolute value detector, 1
Reference numeral 5 is an NM bit data conversion unit for reading out data of one word stored in the A / D data storage memory 13 and having N bits, and converting one word into data having M bits. Linear PCM M-bit word input voice encoder for inputting and encoding voice data, 17 is a code memory for accumulating the encoded voice data, 18 is a linear encoder of M-bit length for the encoded voice data M bit length word output voice decoder for decoding into PCM voice data, 19 is a digital decoder with M bit resolution
An M-bit D / A converter for converting data into an analog signal, 20 is an audio output terminal for outputting an audio signal, 21
Reference numeral 22 denotes an operation unit that has keys and receives various commands when an operator operates, and 22 is a control unit that integrally controls the above-described functional units.

The operation of the voice sampler configured as above will be described below. First, the operation at the time of voice sampling / encoding will be described. First, when a voice sampling start request is made through the operation unit 21, the control unit 22 activates the N-bit A / D converter 12 and receives the input. The audio signal is converted into digital data by the N-bit A / D converter 12 with N-bit resolution and sequentially stored in the A / D data storage memory 13. This operation continues until the next voice sampling stop request is issued via the operation unit 21.

When a voice sampling stop request is issued via the operation unit 21, the control unit 22 interrupts the voice sampling operation, and then activates the maximum absolute value detection unit 14 for A / A conversion. The maximum absolute value in the sampling data accumulated in the D data storage memory 13 is detected. Maximum value MAXn of the absolute value detected here
Based on, when the maximum absolute value that can be represented by M bits is MAXm, the control unit 12 calculates a value of MAXm / MAXn (hereinafter, this value is referred to as MUL), and calculates NM (bits). The MUL is set in the data conversion unit 15 and activated, and at the same time, the M-bit long word input speech encoder 16 is activated.

When the NM bit data conversion unit 15 is activated, A / D data of 1 word N bit width is sequentially read from the A / D data storage memory 13 word by word, and M
UL-multiplied M-bit word input speech coder 16
Entered in. M-bit long word input speech coder 16
Then, the encoded data generated by encoding the audio with the input of the linear A / D data having the M bit width is stored in the code memory 17.

As described above, the voice is sampled and coded. FIG. 2 shows a conceptual diagram of the audio data before and after conversion in the N-M bit data conversion unit 15 at this time with an audio waveform.

Next, the operation at the time of decoding and reproducing the voice will be described. When a voice reproduction start request is issued via the operation unit 21, the control unit 22 activates the M-bit long word output voice decoder 18 and the M-bit D / A converter 19, and is stored in the code memory 17. The encoded voice data is read out, enters the M-bit long word output voice decoder 18, is decoded into M-bit linear data, and is then D / A converted by the M-bit D / A converter 19. The voice is reproduced.

Here, M and N are positive integers, and M <
N and M bit word input speech coder 16
And the decoding algorithm of the M-bit long word output speech decoder 18 is, for example, ADPCM.
The method is not limited to this method as long as the linear PCM audio data processing device is used as an input / output, and a PCM method of storing linear PCM audio data itself in a code memory is also included.

Further, in this embodiment, the conversion of the audio data from N bits to M bits is performed by multiplying the N bit data by the value MUL = MAXm / MAXn. It is also possible to shift each N-bit data to the right by arithmetic calculation so that the position of the most significant bit of a certain bit) becomes the most significant bit position of the maximum positive value in the 2's complement representation of M bits. Is.

FIGS. 3 and 4 are views showing an embodiment of a voice sampling device according to the second invention of the present application. In FIG.
Reference numeral 31 is an audio input terminal for inputting an audio signal, 22 is an N-bit A / D converter which is a means for converting an analog signal of the input audio into digital data with N-bit resolution, and 33 is an N-bit A / D converter. A / D data storage memory for storing voice data converted into digital data by the D converter 32 with N-bit resolution, and 34 for detecting the maximum value in the data stored in the A / D data storage memory 33. A maximum value detecting unit 35 for detecting the minimum value in the data stored in the A / D data storage memory 33, and a minimum value detecting unit 36 for adding a constant to the input data. Is a constant adder for outputting, and 37 is a direct current removing digital filter for performing a direct current removing digital filtering process on the input data string, and 38 is input data 1 performs a process of multiplying a constant for
A constant multiplier that converts the effective bit length of a word from N bits to M bits and outputs the result. Further, reference numeral 39 is an M-bit length word input voice encoder for inputting M-bit length linear PCM voice data, 40 is a code memory for accumulating the encoded voice data, and 41 is an encoder. M-bit word output voice decoder for decoding the voice data into M-bit linear PCM voice data, 42 is an M-bit D / A converter for converting M-bit resolution digital data into an analog signal, Reference numeral 43 is a voice output terminal for outputting a voice signal, 44 is an operation unit which has a key and receives various commands by receiving an operation of an operator, and 45 is a control unit which integrally controls the above functional units. .

The operation of the voice sampler configured as described above will be described below. First, the operation at the time of voice sampling / encoding will be described. First, when a voice sampling start request is made via the operation unit 44, the control unit 45 activates the N-bit A / D converter 32 and inputs voice. The audio signal input from the terminal 31 is converted into digital data by the N-bit A / D converter 32 with N-bit resolution and sequentially stored in the A / D data storage memory 33. This operation continues until the next voice sampling stop request is issued via the operation unit 44.

When a voice sampling stop request is issued via the operation unit 44, the control unit 45 interrupts the voice sampling operation, and then the maximum value detection unit 34.
Also, the minimum value detection unit 35 is activated to detect the maximum value and the minimum value in the sampling data accumulated in the A / D data storage memory 33. Maximum value M detected here
Based on AXn and the minimum value MINn, the control unit 45 calculates OFFSET = (MAXn + MINn) / 2 and sets -OFFSET to the constant adder 36. Further, when the maximum absolute value that can be represented by M bits is MAXm, the value of MUL = MAXm / (MAXn-OFFSET) is calculated and MUL is set in the constant multiplier 38. After that, the control unit 45 controls the constant adder 36, the direct-current removal digital filter 37, the constant multiplier 38, and M.
The bit length word input speech coder 39 is activated.

When the constant adder 36, the DC removal digital filter 37, the constant multiplier 38, and the M-bit-length word input speech encoder 39 are activated, the A / D data storage memory 33 stores one word N-bit width. A / D data of 1
Words are sequentially read and entered into the constant adder 36, −
OFFSET is added. Next, after the DC component is removed through the DC removal digital filter 37, it is MUL-multiplied by the constant multiplier 38 to convert the effective bits of one word from N bits to M bits, and then the M bit long word input voice. It is input to the encoder 39. In the M-bit long word input speech coder 39, a linear A / D having an M-bit width is used.
Encoded data generated by encoding voice with data as input is stored in the code memory 40.

As described above, the voice is sampled and encoded. A voice waveform is a conceptual diagram of the voice data (before conversion) on the A / D data storage memory 33 at this time (before conversion) and the voice data (after conversion) input to the M-bit long word input voice encoder 39. It is FIG.

Next, the operation at the time of decoding and reproducing the voice will be described. When a voice reproduction start request is issued via the operation unit 44, the control unit 45 activates the M-bit long word output voice decoder 41 and the M-bit D / A converter 42 and is stored in the code memory 40. The voice code data is read out, enters the M-bit length word output voice decoder 41, is decoded into M-bit linear data, and is D / A converted by the M-bit D / A converter 42 to output the voice. Is played.

Here, regarding the relationship between M and N, and the algorithms of the M-bit long word input speech encoder and the M-bit long word output speech decoder 18, they are the same as those described in the first embodiment. Is.

[0025]

As described above, according to the present invention,
When voice is input, the voice data is sampled by an A / D converter having a resolution higher than the resolution of input / output data in the voice encoder and the voice decoder, temporarily stored in a memory, and then input to the voice encoder. Since the data resolution is converted to a resolution suitable for the voice encoder in S, even if the input level is set with a margin for a voice input signal with a large amplitude, the S / The N ratio is not lowered. In addition, the voice signal can be sampled by making full use of the dynamic range of the resolution of the input / output data in the voice encoder and the voice decoder and without causing distortion.

In the second aspect of the invention, similarly to the above, after the voice data is stored in the A / D data storage memory, the voice data is calculated so that the maximum amplitude is the same in the + direction and the-direction. After that, the dynamic range of the resolution of the input and output data in the speech coder and the speech decoder is converted by converting the resolution of the data into the resolution suitable for the speech coder at the stage of inputting to the speech coder. Since the voice signal can be sampled by making full use of both + and − directions and without causing distortion, the S / N ratio is large even when the difference between the + direction amplitude and the − direction amplitude of the input voice is large. The effect that the audio signal can be sampled without lowering the ratio or causing distortion is obtained.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a voice sampling device according to the first invention of the present application.

FIG. 2 is a conceptual diagram showing audio data before and after NM bit data conversion in the embodiment.

FIG. 3 is a block diagram showing an embodiment of a voice sampling device according to the second invention of the present application.

FIG. 4 is a conceptual diagram of voice data (before conversion) on an A / D data storage memory and voice data (after conversion) input to an M-bit word input voice encoder in the above embodiment.

FIG. 5 is a block diagram showing an example of a conventional voice sampling device.

[Explanation of symbols]

 11, 31 voice input terminal 12, 32 N-bit A / D converter 13, 33 A / D data storage memory 14 maximum absolute value detection unit 15 NM bit data conversion unit 16, 39 M-bit long word input voice encoding Device 17, 40 Code memory 18, 41 M-bit long word output audio decoder 19, 42 M-bit D / A converter 20, 43 Audio output terminal 21, 44 Operation unit 22, 45 Control unit 34 Maximum value detection unit 35 Minimum value detector 36 Constant adder 37 DC removal digital filter 38 Constant multiplier

Claims (2)

[Claims] 1. A means for A / D converting an analog audio signal with a resolution of N bits, and a storage means for storing data of a resolution of N bits output from the A / D conversion means.
Data from N-bit resolution to M-bit resolution by multiplying N-bit resolution data by a voice encoding means that inputs data after A / D conversion with M-bit (M <N) resolution A resolution conversion means for performing conversion, and A / D-converts the input audio signal with a resolution of N bits, temporarily stores it in the storage means, and then performs resolution conversion of data from N bits to M bits by the resolution conversion means. A voice sample device characterized in that the voice signal is encoded by the voice encoding means. 2. Means for A / D converting an analog audio signal with N-bit resolution, and storage means for storing data of N-bit resolution output from this A / D conversion means.
A voice encoding means for inputting data after A / D conversion having a resolution of M bits (M <N), and a maximum amplitude in the ± direction of the amplitude with respect to the data of the N bit resolution stored in the storage means. Audio data conversion means for performing data conversion from N-bit resolution to M-bit resolution after conversion so that the input audio signal has an A-bit resolution of A.
A voice sampling device characterized in that after D / D conversion, it is temporarily stored in a storage means, the voice data is converted by the voice data converting means, and then a voice signal is encoded by the voice encoding means.